Vacation Package Project

By: Kyle Bowen & Will Kittredge

Objective

A travel agency has asked you and your team to create a machine learning model to help them predict whether or not potential customers will purchase a new vacation package they are offering. As part of this process, the agency would also like to understand the potential customers better to more efficiently market their new package.

Data Definition

• CustomerID: Unique customer ID
• ProdTaken: Whether the customer has purchased a package or not (0: No, 1: Yes)
• Age: Age of customer
• TypeofContact: How customer was contacted (Company Invited or Self Inquiry)
• CityTier: City tier depends on the development of a city, population, facilities, and living standards. The categories are ordered i.e. Tier 1 > Tier 2 > Tier 3
• DurationOfPitch: Duration of the pitch by a salesperson to the customer
• Occupation: Occupation of customer
• Gender: Gender of customer
• NumberOfPersonVisiting: Total number of people planning to take the trip with the customer
• NumberOfFollowups: Total number of follow-ups has been done by the salesperson after the sales pitch
• ProductPitched: Product pitched by the salesperson
• PreferredPropertyStar: Preferred hotel property rating by customer
• MaritalStatus: Marital status of customer
• NumberOfTrips: Average number of trips in a year by customer
• Passport: The customer has a passport or not (0: No, 1: Yes)
• PitchSatisfactionScore: Sales pitch satisfaction score
• OwnCar: Whether the customers own a car or not (0: No, 1: Yes)
• NumberOfChildrenVisiting: Total number of children with age less than 5 planning to take the trip with the customer
• Designation: Designation of the customer in the current organization
• MonthlyIncome: Gross monthly income of the customer

Boilerplate

import pandas as pd
import numpy as np


import matplotlib.pyplot as plt
import seaborn as sns

import sklearn.preprocessing
scaler = sklearn.preprocessing.MinMaxScaler()
labelEncoder = sklearn.preprocessing.LabelEncoder()

sns.set(color_codes=True)
%matplotlib inline

df = pd.read_csv("vacation_package.csv")

Functions

def ReturnByPurchase(columnparam):
    purchased = df.loc[(df[columnparam].isnull() == False) & (df['ProdTaken'] == 1)]
    nopurchase = df.loc[(df[columnparam].isnull() == False) & (df['ProdTaken'] == 0)]
    
    print("Purchased Mean: " + str(round(purchased[columnparam].mean(),6)))
    print("Purchased Median: " + str(purchased[columnparam].median()))
    print("Purchased Mode: " + str(purchased[columnparam].mode()))
    print('\n' + "-----------" + '\n')
    print("Unpurchased Mean: " + str(round(nopurchase[columnparam].mean(),6)))
    print("Unpurchased Median: " + str(nopurchase[columnparam].median()))
    print("Unpurchased Mode: " + str(nopurchase[columnparam].mode()))

def histogramboxplot(data, feature, figsize=(12,7), kde=True, bins=None):
    """
    Boxplot and histogram combined

    data: dataframe
    feature: dataframe column
    figsize: size of figure (default (12,7))
    kde: whether to show the density curve (default False)
    bins: number of bins for histogram (default None)
    """
    f2, (ax_box2, ax_hist2) = plt.subplots(
        nrows=2,  # Number of rows of the subplot grid= 2
        sharex=True,  # x-axis will be shared among all subplots
        gridspec_kw={"height_ratios": (0.25, 0.75)},
        figsize=figsize,
    )  # creating the 2 subplots
    
    sns.boxplot(
        data=data, x=feature, ax=ax_box2, showmeans=True, meanprops = dict(marker='D', markeredgecolor='black', markerfacecolor='red')
    )  # boxplot will be created and a star will indicate the mean value of the column
    
    sns.histplot(
        data=data, x=feature, kde=kde, ax=ax_hist2, bins=bins
    ) if bins else sns.histplot(
        data=data, x=feature, kde=kde, ax=ax_hist2
    )  # For histogram
    
    ax_hist2.axvline(
        data[feature].mean(), color="red", linestyle="-"
    )  # Add mean to the histogram
    
    ax_hist2.axvline(
        data[feature].median(), color="black", linestyle="--"
    )  # Add median to the histogram

def stackedbarplot(data, predictor, target):
    """
    Print the category counts and plot a stacked bar chart

    data: dataframe
    predictor: independent variable
    target: target variable
    """
    count = data[predictor].nunique()
    sorter = data[target].value_counts().index[-1]
    tab1 = pd.crosstab(data[predictor], data[target], margins=True).sort_values(
        by=sorter, ascending=False
    )
    print(tab1)
    print("-" * 115)
    tab = pd.crosstab(data[predictor], data[target], normalize="index").sort_values(
        by=sorter, ascending=False
    )
    print(tab)
    tab.plot(kind="bar", stacked=True, figsize=(count + 5, 5))
    plt.legend(
        loc="lower left", frameon=False,
    )
    plt.legend(loc="upper left", bbox_to_anchor=(1, 1))
    plt.show()

def descviolinbox(Data, tablename):
    """
    Prints the Description of a numeric value and a stacked violin and boxplot 
    
    Data = Dataframe 
    
    value = Target Column
    
    """
    tablename = str(tablename)
    
    description = Data[tablename].describe()
    
    sns.violinplot(data=Data, y=tablename)

    sns.boxplot(data=Data, y=tablename, color="gray")
    
    return description

def bounds(Data, tablename):
    """
    Returns upper and lower bounds for a table, helps us capture outliers
    
    Data = Dataframe 
    tablename = Target Column
    """
    tablename = str(tablename)
    
    Q1 = Data[tablename].quantile(0.25)
    Q3 = Data[tablename].quantile(0.75)
    IQR = Q3 - Q1
    lowerlim = Q1 - 1.5 * IQR
    upperlim = Q3 + 1.5 * IQR
    
    print("Q1: " + str(Q1))
    print("Q3: " + str(Q3))
    print("IQR: " + str(IQR))
    print("lower limit: " + str(lowerlim))
    print("upper limit: " + str(upperlim))
    
    return(lowerlim, upperlim, Q1, Q3)

def numericdata(Data, tablename):
    """
    Returns the Mean, Median, Mode, and Range for a dataset
    
    Data = Dataframe
    tablename = target column
    """
    tablename = str(tablename)
    
    Q1 = Data[tablename].quantile(0.25)
    Q3 = Data[tablename].quantile(0.75)
    IQR = Q3-Q1
    mean = Data[tablename].mean()
    median = Data[tablename].median()
    mode = Data[tablename].mode()[0]
    
    mean = round(mean,6)
    median = round(median,6)
    mode = round(mode,6)
    IQR = round(IQR,6)
 
    print("Mean: " + str(mean))
    print("Median: " + str(median))
    print("Mode: " + str(mode))
    print("Range: " + str(IQR))

def labeldisplot(Data, tablename): 
    
        tablename = str(tablename)
        
        sns.displot(data=Data, x=tablename, kde=True);
        plt.axvline(x=Data[tablename].median(), color = 'black')
        plt.axvline(x=Data[tablename].mean(), color = 'red', ls='--')

def labeledbarplot(data, feature, perc=False, n=None):
    """
    Barplot with percentage at the top

    data: dataframe
    feature: dataframe column
    perc: whether to display percentages instead of count (default is False)
    n: displays the top n category levels (default is None, i.e., display all levels)
    """

    total = len(data[feature])  # length of the column
    count = data[feature].nunique()
    if n is None:
        plt.figure(figsize=(count + 1, 5))
    else:
        plt.figure(figsize=(n + 1, 5))

    plt.xticks(rotation=90, fontsize=15)
    ax = sns.countplot(
        data=data,
        x=feature,
        palette="Paired",
        order=data[feature].value_counts().index[:n].sort_values(),
    )

    for p in ax.patches:
        if perc == True:
            label = "{:.1f}%".format(
                100 * p.get_height() / total
            )  # percentage of each class of the category
        else:
            label = p.get_height()  # count of each level of the category

        x = p.get_x() + p.get_width() / 2  # width of the plot
        y = p.get_height()  # height of the plot

        ax.annotate(
            label,
            (x, y),
            ha="center",
            va="center",
            size=12,
            xytext=(0, 5),
            textcoords="offset points",
        )  # annotate the percentage

    plt.show()  # show the plot

def stackedbox(data, value, hue): 
    """
    data= Dataframe
    Value = Independent variable we want to test 
    Hue = Dependent Variable we want to include
    """
    sns.catplot(data= data, x = hue, y = value, kind="box")

def minmeanmax(data, independent, dependent):
    """
    Prints the min, mean and max of things grouped by our dependent variable 
    
    data = Dataframe 
    independent = The variable we want to do math on 
    Dependent = the variable we want to group by 
    """
    
    notpurchased = df.loc[df[dependent] == 0] #Adjust Number of df.Locs for number of variables
    purchased = df.loc[df[dependent] == 1]
    
    print("Not Purchased Min Mean and Max:" + '\n')
    print(f' Min: {notpurchased[independent].min()}')
    print(f' Mean: {round(notpurchased[independent].mean(), 6)}')
    print(f' Max: {notpurchased[independent].max()}')
    
    print("\n" + "Purchased Min Mean and Max:" + "\n")
    print(f' Min: {purchased[independent].min()}')  
    print(f' Mean: {round(purchased[independent].mean(), 6)}') 
    print(f' Max: {purchased[independent].max()}')
    
    #Add as many print statement blocks as required for the dependent variable, keep this count low

Data Overview

# df.sample(10)

	CustomerID	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
996	200996	0	27.0	Self Enquiry	3	16.0	Small Business	Female	3	4.0	Deluxe	3.0	Divorced	2.0	1	3	1	2.0	Manager	20769.0
100	200100	1	37.0	Self Enquiry	2	12.0	Salaried	Male	3	3.0	Basic	5.0	Married	5.0	1	2	1	0.0	Executive	17073.0
4493	204493	0	35.0	Self Enquiry	1	9.0	Small Business	Female	3	5.0	Basic	5.0	Unmarried	3.0	0	1	1	1.0	Executive	23059.0
3579	203579	0	47.0	Self Enquiry	3	7.0	Salaried	Male	3	2.0	Super Deluxe	5.0	Single	NaN	0	4	1	2.0	AVP	36245.0
320	200320	0	27.0	Self Enquiry	3	NaN	Salaried	Male	3	3.0	Deluxe	3.0	Single	2.0	1	4	1	2.0	Manager	NaN
2585	202585	0	46.0	Self Enquiry	1	36.0	Small Business	Male	3	4.0	Basic	3.0	Unmarried	7.0	0	2	1	1.0	Executive	22130.0
3409	203409	0	26.0	Self Enquiry	1	26.0	Small Business	Male	4	4.0	Basic	3.0	Divorced	5.0	0	5	1	3.0	Executive	22347.0
3718	203718	0	32.0	Self Enquiry	3	36.0	Small Business	Female	4	5.0	Deluxe	3.0	Married	3.0	0	3	1	1.0	Manager	24146.0
2039	202039	0	36.0	Company Invited	3	14.0	Salaried	Male	3	4.0	Standard	5.0	Unmarried	2.0	0	1	0	0.0	Senior Manager	22587.0
3581	203581	0	23.0	Company Invited	1	33.0	Salaried	Female	3	5.0	Basic	3.0	Married	3.0	1	3	0	1.0	Executive	21492.0

# df.sample(10)

	CustomerID	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
3602	203602	0	38.0	Self Enquiry	1	26.0	Salaried	Male	3	4.0	Deluxe	3.0	Married	5.0	0	1	0	1.0	Manager	24446.0
2235	202235	0	38.0	Company Invited	1	9.0	Salaried	Male	2	3.0	Basic	3.0	Married	4.0	0	3	1	0.0	Executive	17821.0
4820	204820	1	35.0	Self Enquiry	1	14.0	Salaried	Male	4	4.0	Deluxe	3.0	Married	3.0	0	3	0	1.0	Manager	21263.0
2864	202864	0	30.0	Self Enquiry	1	10.0	Small Business	Male	4	5.0	Standard	3.0	Divorced	3.0	0	2	1	3.0	Senior Manager	30613.0
1577	201577	1	25.0	Self Enquiry	3	11.0	Small Business	Male	2	4.0	Deluxe	3.0	Single	2.0	1	3	0	1.0	Manager	20744.0
3342	203342	0	44.0	Self Enquiry	1	10.0	Salaried	Male	4	6.0	King	NaN	Divorced	5.0	0	5	1	3.0	VP	38418.0
1303	201303	0	47.0	Self Enquiry	1	10.0	Salaried	Female	3	4.0	Standard	4.0	Married	1.0	1	4	1	2.0	Senior Manager	25333.0
666	200666	1	NaN	Self Enquiry	1	9.0	Salaried	Female	2	3.0	Deluxe	3.0	Divorced	1.0	1	5	1	1.0	Manager	NaN
1129	201129	0	42.0	Self Enquiry	1	15.0	Salaried	Male	3	4.0	King	3.0	Single	1.0	0	1	1	1.0	VP	34613.0
1232	201232	0	35.0	Self Enquiry	1	33.0	Salaried	Male	2	3.0	Deluxe	3.0	Married	3.0	0	5	0	1.0	Manager	21883.0

Observation

Null values appear to be represented as NaN

df.shape

(4888, 20)

Observation

There appears to be 20 columns and 4888 rows

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4888 entries, 0 to 4887
Data columns (total 20 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   CustomerID                4888 non-null   int64  
 1   ProdTaken                 4888 non-null   int64  
 2   Age                       4662 non-null   float64
 3   TypeofContact             4863 non-null   object 
 4   CityTier                  4888 non-null   int64  
 5   DurationOfPitch           4637 non-null   float64
 6   Occupation                4888 non-null   object 
 7   Gender                    4888 non-null   object 
 8   NumberOfPersonVisiting    4888 non-null   int64  
 9   NumberOfFollowups         4843 non-null   float64
 10  ProductPitched            4888 non-null   object 
 11  PreferredPropertyStar     4862 non-null   float64
 12  MaritalStatus             4888 non-null   object 
 13  NumberOfTrips             4748 non-null   float64
 14  Passport                  4888 non-null   int64  
 15  PitchSatisfactionScore    4888 non-null   int64  
 16  OwnCar                    4888 non-null   int64  
 17  NumberOfChildrenVisiting  4822 non-null   float64
 18  Designation               4888 non-null   object 
 19  MonthlyIncome             4655 non-null   float64
dtypes: float64(7), int64(7), object(6)
memory usage: 763.9+ KB

Observation

Our data appears to all be floats, integers, and objects, and they all make sense for where they are. There appear to be a few null values in our dataset

df.isnull().sum()

CustomerID                    0
ProdTaken                     0
Age                         226
TypeofContact                25
CityTier                      0
DurationOfPitch             251
Occupation                    0
Gender                        0
NumberOfPersonVisiting        0
NumberOfFollowups            45
ProductPitched                0
PreferredPropertyStar        26
MaritalStatus                 0
NumberOfTrips               140
Passport                      0
PitchSatisfactionScore        0
OwnCar                        0
NumberOfChildrenVisiting     66
Designation                   0
MonthlyIncome               233
dtype: int64

Observation

we have welll over 200 empty rows in three columns, one column with over 100 and 4 more with less than 100, these will need to be handled

df.duplicated().sum()

0

Observation

There are no duplicate rows in our dataset

Exploring Numeric Data

df.describe().T

	count	mean	std	min	25%	50%	75%	max
CustomerID	4888.0	202443.500000	1411.188388	200000.0	201221.75	202443.5	203665.25	204887.0
ProdTaken	4888.0	0.188216	0.390925	0.0	0.00	0.0	0.00	1.0
Age	4662.0	37.622265	9.316387	18.0	31.00	36.0	44.00	61.0
CityTier	4888.0	1.654255	0.916583	1.0	1.00	1.0	3.00	3.0
DurationOfPitch	4637.0	15.490835	8.519643	5.0	9.00	13.0	20.00	127.0
NumberOfPersonVisiting	4888.0	2.905074	0.724891	1.0	2.00	3.0	3.00	5.0
NumberOfFollowups	4843.0	3.708445	1.002509	1.0	3.00	4.0	4.00	6.0
PreferredPropertyStar	4862.0	3.581037	0.798009	3.0	3.00	3.0	4.00	5.0
NumberOfTrips	4748.0	3.236521	1.849019	1.0	2.00	3.0	4.00	22.0
Passport	4888.0	0.290917	0.454232	0.0	0.00	0.0	1.00	1.0
PitchSatisfactionScore	4888.0	3.078151	1.365792	1.0	2.00	3.0	4.00	5.0
OwnCar	4888.0	0.620295	0.485363	0.0	0.00	1.0	1.00	1.0
NumberOfChildrenVisiting	4822.0	1.187267	0.857861	0.0	1.00	1.0	2.00	3.0
MonthlyIncome	4655.0	23619.853491	5380.698361	1000.0	20346.00	22347.0	25571.00	98678.0

# df.mode()

	CustomerID	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	200000	0.0	35.0	Self Enquiry	1.0	9.0	Salaried	Male	3.0	4.0	Basic	3.0	Married	2.0	0.0	3.0	1.0	1.0	Executive	17342.0
1	200001	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	20855.0
2	200002	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	21020.0
3	200003	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	21288.0
4	200004	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4883	204883	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4884	204884	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4885	204885	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4886	204886	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4887	204887	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

4888 rows × 20 columns

Observations

• Most people did not purchase a package
• Ages range from 18 to 61, with the middle being in the mid 30s
• City Tier ranges from 1-3, with most people being in a Tier 1 city
• Pitches ranged from 5 minutes to 2 hours with most being 13-15 minutes
• Only 1-5 people visited a destination, with most common number of people visiting being three
• There were usually 4 followups
• The most frequent product pitched was the basic package
• Preferred Properties ranged from 3-5 stars, with most being 3 stars
• Most individuals wer married
• There were between 1 to 22 trips taken annually, with the middle being 3, and most people taking 2.
• Most people did not have a passport
• Pitch Satisfaction Scores hovered around 3
• Most Individuals owned a car
• On average one child went with each person
• The most common designation was executive
• The average monthly salary hovered between 22 and 23k

# df.corr()

	CustomerID	ProdTaken	Age	CityTier	DurationOfPitch	NumberOfPersonVisiting	NumberOfFollowups	PreferredPropertyStar	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	MonthlyIncome
CustomerID	1.000000	0.056506	0.032127	0.012975	0.064298	0.604013	0.427539	0.010553	0.224848	0.007974	-0.035847	0.003805	0.511763	0.276833
ProdTaken	0.056506	1.000000	-0.147254	0.086852	0.078257	0.009627	0.112171	0.099577	0.018898	0.260844	0.051394	-0.011508	0.007421	-0.130585
Age	0.032127	-0.147254	1.000000	-0.015625	-0.012063	0.011621	-0.002577	-0.010474	0.184905	0.033399	0.018510	0.048654	0.007370	0.464869
CityTier	0.012975	0.086852	-0.015625	1.000000	0.022703	-0.001671	0.023652	-0.009164	-0.029709	0.001793	-0.042160	0.003817	0.000672	0.051817
DurationOfPitch	0.064298	0.078257	-0.012063	0.022703	1.000000	0.065141	0.009434	-0.006637	0.009715	0.033034	-0.002880	-0.001626	0.031408	-0.006252
NumberOfPersonVisiting	0.604013	0.009627	0.011621	-0.001671	0.065141	1.000000	0.328569	0.033867	0.195223	0.011177	-0.019581	0.010362	0.610621	0.195134
NumberOfFollowups	0.427539	0.112171	-0.002577	0.023652	0.009434	0.328569	1.000000	-0.024176	0.139517	0.004970	0.004054	0.012112	0.286425	0.176503
PreferredPropertyStar	0.010553	0.099577	-0.010474	-0.009164	-0.006637	0.033867	-0.024176	1.000000	0.012115	0.001040	-0.022701	0.015742	0.035798	0.014289
NumberOfTrips	0.224848	0.018898	0.184905	-0.029709	0.009715	0.195223	0.139517	0.012115	1.000000	0.012949	-0.004378	-0.011825	0.168795	0.139105
Passport	0.007974	0.260844	0.033399	0.001793	0.033034	0.011177	0.004970	0.001040	0.012949	1.000000	0.002926	-0.022330	0.020264	0.002545
PitchSatisfactionScore	-0.035847	0.051394	0.018510	-0.042160	-0.002880	-0.019581	0.004054	-0.022701	-0.004378	0.002926	1.000000	0.068850	0.000878	0.030421
OwnCar	0.003805	-0.011508	0.048654	0.003817	-0.001626	0.010362	0.012112	0.015742	-0.011825	-0.022330	0.068850	1.000000	0.026572	0.080262
NumberOfChildrenVisiting	0.511763	0.007421	0.007370	0.000672	0.031408	0.610621	0.286425	0.035798	0.168795	0.020264	0.000878	0.026572	1.000000	0.201643
MonthlyIncome	0.276833	-0.130585	0.464869	0.051817	-0.006252	0.195134	0.176503	0.014289	0.139105	0.002545	0.030421	0.080262	0.201643	1.000000

Observations

Most correlations in this dataset are weak with ProdTaken, most heavily with NumberofFollowups, and Passport being relatively positively correlated, and age being moderately negatively correlated. Some data changes will need to be done

Checking for Odd Categorical Data

df.nunique()

CustomerID                  4888
ProdTaken                      2
Age                           44
TypeofContact                  2
CityTier                       3
DurationOfPitch               34
Occupation                     4
Gender                         3
NumberOfPersonVisiting         5
NumberOfFollowups              6
ProductPitched                 5
PreferredPropertyStar          3
MaritalStatus                  4
NumberOfTrips                 12
Passport                       2
PitchSatisfactionScore         5
OwnCar                         2
NumberOfChildrenVisiting       4
Designation                    5
MonthlyIncome               2475
dtype: int64

Observations

There are some odd things here, like three unique genders and four unique marital statuses.

df['Gender'].value_counts()

Male       2916
Female     1817
Fe Male     155
Name: Gender, dtype: int64

There are 155 incorrect labels for the "Female" Category

df['MaritalStatus'].value_counts()

Married      2340
Divorced      950
Single        916
Unmarried     682
Name: MaritalStatus, dtype: int64

There isn't anything wrong with this category, but it might be possible to lump them int married and unmarried if we want to down the line

Removing Unnecessary Column(s)

We know each value is unique, but we do not need an ID, since we do not desire to keep track of individual customers in our dataset, rather we want to be able to predict based on values that have bearing besides autoincrements

df = df.drop(['CustomerID'], axis = 1)

# df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3	3.0	Deluxe	3.0	Single	1.0	1	2	1	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	3	4.0	Deluxe	4.0	Divorced	2.0	0	3	1	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	2	3.0	Basic	3.0	Divorced	2.0	1	5	1	1.0	Executive	17909.0
4	0	NaN	Self Enquiry	1	8.0	Small Business	Male	2	3.0	Basic	4.0	Divorced	1.0	0	5	1	0.0	Executive	18468.0

Data Preprocessing

Handling Null values

df.isnull().sum()

ProdTaken                     0
Age                         226
TypeofContact                25
CityTier                      0
DurationOfPitch             251
Occupation                    0
Gender                        0
NumberOfPersonVisiting        0
NumberOfFollowups            45
ProductPitched                0
PreferredPropertyStar        26
MaritalStatus                 0
NumberOfTrips               140
Passport                      0
PitchSatisfactionScore        0
OwnCar                        0
NumberOfChildrenVisiting     66
Designation                   0
MonthlyIncome               233
dtype: int64

Age

df.Age.isna().sum()

226

df.Age.describe().T

count    4662.000000
mean       37.622265
std         9.316387
min        18.000000
25%        31.000000
50%        36.000000
75%        44.000000
max        61.000000
Name: Age, dtype: float64

sns.displot(df['Age'], kde = True)
plt.axvline(x = df.Age.median(), color = 'black', ls='--')
plt.axvline(x = df.Age.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x249363edb20>

Observations

The data is skewed right, and given that it likely has a bearing on the dependent variable, and a difference depending on whether or not they purchased, imputing the median based on whether or not they purchased ticket seems like a smart idea.

Finding the values

ReturnByPurchase('Age')

Purchased Mean: 34.770548
Purchased Median: 33.0
Purchased Mode: 0    29.0
Name: Age, dtype: float64

-----------

Unpurchased Mean: 38.282092
Unpurchased Median: 37.0
Unpurchased Mode: 0    36.0
Name: Age, dtype: float64

Setting the Values

df.loc[(df['Age'].isnull() == True) & (df['ProdTaken'] == 1), ['Age']] = 33
df.loc[(df['Age'].isnull() == True) & (df['ProdTaken'] == 0),['Age']] = 37

Examining Changes

df.Age.isnull().sum()

0

df.Age.describe().T

count    4888.000000
mean       37.557488
std         9.109545
min        18.000000
25%        31.000000
50%        37.000000
75%        43.000000
max        61.000000
Name: Age, dtype: float64

sns.displot(df['Age'], kde = True)
plt.axvline(x = df.Age.median(), color = 'black', ls='--')
plt.axvline(x = df.Age.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x24937bbf520>

The median and the mean appear to have been brought closer together in this dataset due to the imputing we just did, but overall the datas structure has not changed much

Type of Contact

df.TypeofContact.isna().sum()

25

# df[df['TypeofContact'].isnull()]

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
224	0	31.0	NaN	1	NaN	Small Business	Male	2	5.0	Deluxe	3.0	Divorced	1.0	0	3	1	0.0	Manager	NaN
571	0	26.0	NaN	1	NaN	Salaried	Female	3	5.0	Basic	3.0	Married	4.0	0	4	1	2.0	Executive	NaN
572	0	29.0	NaN	1	NaN	Small Business	Female	3	3.0	Deluxe	3.0	Divorced	5.0	0	2	1	0.0	Manager	NaN
576	0	27.0	NaN	3	NaN	Small Business	Male	2	3.0	Deluxe	3.0	Divorced	1.0	0	3	0	1.0	Manager	NaN
579	0	34.0	NaN	1	NaN	Small Business	Female	2	4.0	Basic	5.0	Single	2.0	0	2	1	1.0	Executive	NaN
598	1	28.0	NaN	1	NaN	Small Business	Male	2	3.0	Basic	3.0	Single	7.0	0	3	0	0.0	Executive	NaN
622	0	32.0	NaN	3	NaN	Salaried	Male	3	3.0	Deluxe	3.0	Married	3.0	0	2	0	0.0	Manager	NaN
724	0	24.0	NaN	1	NaN	Small Business	Female	2	4.0	Deluxe	3.0	Married	2.0	0	3	1	1.0	Manager	NaN
843	0	26.0	NaN	1	NaN	Small Business	Male	2	1.0	Basic	3.0	Divorced	2.0	0	5	1	1.0	Executive	NaN
1021	1	25.0	NaN	3	NaN	Salaried	Male	3	4.0	Basic	5.0	Divorced	4.0	0	1	1	0.0	Executive	NaN
1047	0	33.0	NaN	3	NaN	Small Business	Male	2	3.0	Deluxe	5.0	Divorced	1.0	0	3	0	0.0	Manager	NaN
1143	0	45.0	NaN	3	NaN	Small Business	Male	2	4.0	Deluxe	5.0	Married	2.0	0	3	0	0.0	Manager	NaN
1182	0	36.0	NaN	1	NaN	Small Business	Female	2	4.0	Deluxe	3.0	Married	1.0	0	5	1	1.0	Manager	NaN
1217	0	24.0	NaN	1	NaN	Small Business	Male	3	1.0	Basic	3.0	Married	2.0	0	1	0	0.0	Executive	NaN
1356	0	41.0	NaN	3	NaN	Small Business	Female	2	3.0	Deluxe	4.0	Married	6.0	0	3	1	1.0	Manager	NaN
1469	0	34.0	NaN	1	NaN	Small Business	Male	2	1.0	Deluxe	3.0	Married	3.0	0	3	0	1.0	Manager	NaN
1694	0	31.0	NaN	1	NaN	Small Business	Male	2	5.0	Deluxe	3.0	Married	1.0	0	3	0	0.0	Manager	NaN
2041	0	26.0	NaN	1	NaN	Salaried	Female	3	5.0	Basic	3.0	Married	4.0	0	4	1	0.0	Executive	NaN
2042	0	29.0	NaN	1	NaN	Small Business	Female	3	3.0	Deluxe	3.0	Married	5.0	0	1	0	1.0	Manager	NaN
2046	0	27.0	NaN	3	NaN	Small Business	Male	2	3.0	Deluxe	3.0	Married	1.0	0	3	1	1.0	Manager	NaN
2049	0	34.0	NaN	1	NaN	Small Business	Female	2	4.0	Basic	5.0	Single	2.0	0	1	1	0.0	Executive	NaN
2068	1	28.0	NaN	1	NaN	Small Business	Male	2	3.0	Basic	3.0	Single	7.0	0	3	1	1.0	Executive	NaN
2092	0	32.0	NaN	3	NaN	Salaried	Male	3	3.0	Deluxe	3.0	Married	3.0	0	1	0	2.0	Manager	NaN
2194	0	24.0	NaN	1	NaN	Small Business	Female	2	4.0	Deluxe	3.0	Married	2.0	0	3	0	0.0	Manager	NaN
2313	0	26.0	NaN	1	NaN	Small Business	Male	2	1.0	Basic	3.0	Married	2.0	0	5	1	1.0	Executive	NaN

df['TypeofContact'].mode()

0    Self Enquiry
Name: TypeofContact, dtype: object

df.loc[df.TypeofContact.isnull(), 'TypeofContact'] = 'Self Enquiry'

Notes

With only 25 missing fields, and the vast majority of TypeofContact being self enquiry, we just imputed with the mode of the TypeofContact.

df.TypeofContact.isna().sum()

0

Duration of Pitch

df.DurationOfPitch.isna().sum()

251

df.DurationOfPitch.describe()

count    4637.000000
mean       15.490835
std         8.519643
min         5.000000
25%         9.000000
50%        13.000000
75%        20.000000
max       127.000000
Name: DurationOfPitch, dtype: float64

sns.displot(df['DurationOfPitch'], kde = True)
plt.axvline(x = df.DurationOfPitch.median(), color = 'black', ls='--')
plt.axvline(x = df.DurationOfPitch.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x24937bd7ca0>

Observations

This data is extremely skewed right, we will have to come back when we are examining outliers later on and fix this, as I believe the 120 is an inaccurate number, given the KDE near 1 out to 120. We will have to impute with the median rather than the mean for null values. Due to the large number, splitting by whether they purchased or not seems wise again.

Finding the Values

ReturnByPurchase('DurationOfPitch')

Purchased Mean: 16.873143
Purchased Median: 15.0
Purchased Mode: 0    9.0
Name: DurationOfPitch, dtype: float64

-----------

Unpurchased Mean: 15.169325
Unpurchased Median: 13.0
Unpurchased Mode: 0    9.0
Name: DurationOfPitch, dtype: float64

Setting the Values

df.loc[(df['DurationOfPitch'].isnull() == True) & (df['ProdTaken'] == 1), ['DurationOfPitch']] = 15
df.loc[(df['DurationOfPitch'].isnull() == True) & (df['ProdTaken'] == 0),['DurationOfPitch']] = 13

Examining Changes

df.DurationOfPitch.describe()

count    4888.000000
mean       15.381342
std         8.313127
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max       127.000000
Name: DurationOfPitch, dtype: float64

sns.displot(df['DurationOfPitch'], kde = True)
plt.axvline(x = df.DurationOfPitch.median(), color = 'black', ls='--')
plt.axvline(x = df.DurationOfPitch.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x24937e9bfd0>

There appears to be little difference in the structure of the data after these changes, except the increase in the number of objects in the median

Number of Followups

df.NumberOfFollowups.isna().sum()

45

df.NumberOfFollowups.describe()

count    4843.000000
mean        3.708445
std         1.002509
min         1.000000
25%         3.000000
50%         4.000000
75%         4.000000
max         6.000000
Name: NumberOfFollowups, dtype: float64

sns.displot(df['NumberOfFollowups'], kde = True)
plt.axvline(x = df.NumberOfFollowups.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfFollowups.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x24937f74310>

Observations

This is mainly categorical with an integer, with the median and mode both being 4, and the mean a little off at 3.7. This likely has bearing on the dependent variable, so we will impute with that in mind

Finding the Values

ReturnByPurchase('NumberOfFollowups')

Purchased Mean: 3.941886
Purchased Median: 4.0
Purchased Mode: 0    4.0
Name: NumberOfFollowups, dtype: float64

-----------

Unpurchased Mean: 3.654286
Unpurchased Median: 4.0
Unpurchased Mode: 0    4.0
Name: NumberOfFollowups, dtype: float64

Setting the Values

df.loc[df['NumberOfFollowups'].isnull(), ['NumberOfFollowups']] = 4

Checking the Results

df.NumberOfFollowups.isna().sum()

0

df.NumberOfFollowups.describe()

count    4888.000000
mean        3.711129
std         0.998271
min         1.000000
25%         3.000000
50%         4.000000
75%         4.000000
max         6.000000
Name: NumberOfFollowups, dtype: float64

sns.displot(df['NumberOfFollowups'], kde = True)
plt.axvline(x = df.NumberOfFollowups.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfFollowups.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x24937fe0610>

The data looks to mostly be in the same shape as before, due to the low number of changes to fields

Preferred Property Star

df.PreferredPropertyStar.isna().sum()

26

df.PreferredPropertyStar.describe()

count    4862.000000
mean        3.581037
std         0.798009
min         3.000000
25%         3.000000
50%         3.000000
75%         4.000000
max         5.000000
Name: PreferredPropertyStar, dtype: float64

sns.displot(df['PreferredPropertyStar'], kde = True)
plt.axvline(x = df.PreferredPropertyStar.median(), color = 'black', ls='--')
plt.axvline(x = df.PreferredPropertyStar.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x249380501f0>

Observations

This is another categorical value with an integer base. The median and Mode are both at 3, and the mean is a bit above at 3.58 I will impute with the median once more

Finding the Values

ReturnByPurchase('PreferredPropertyStar')

Purchased Mean: 3.746171
Purchased Median: 3.0
Purchased Mode: 0    3.0
Name: PreferredPropertyStar, dtype: float64

-----------

Unpurchased Mean: 3.542806
Unpurchased Median: 3.0
Unpurchased Mode: 0    3.0
Name: PreferredPropertyStar, dtype: float64

df.loc[df['PreferredPropertyStar'].isnull(), ['PreferredPropertyStar']] = 3

Checking the Results

df.PreferredPropertyStar.isna().sum()

0

df.PreferredPropertyStar.describe()

count    4888.000000
mean        3.577946
std         0.797005
min         3.000000
25%         3.000000
50%         3.000000
75%         4.000000
max         5.000000
Name: PreferredPropertyStar, dtype: float64

sns.displot(df['PreferredPropertyStar'], kde = True)
plt.axvline(x = df.PreferredPropertyStar.median(), color = 'black', ls='--')
plt.axvline(x = df.PreferredPropertyStar.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x249380d2f10>

As expected at this point, the data appears to mostly be the same as before.

Number of Trips

df.NumberOfTrips.isna().sum()

140

df.NumberOfTrips.describe()

count    4748.000000
mean        3.236521
std         1.849019
min         1.000000
25%         2.000000
50%         3.000000
75%         4.000000
max        22.000000
Name: NumberOfTrips, dtype: float64

sns.displot(df['NumberOfTrips'], kde = True)
plt.axvline(x = df.NumberOfTrips.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfTrips.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x2493912a880>

Observations

The median and mean are close, and there is definitely an outlier up at 22, given the right skew, I will impute our missing values with the Median

Finding the Values

ReturnByPurchase('NumberOfTrips')

Purchased Mean: 3.30837
Purchased Median: 3.0
Purchased Mode: 0    2.0
Name: NumberOfTrips, dtype: float64

-----------

Unpurchased Mean: 3.219531
Unpurchased Median: 3.0
Unpurchased Mode: 0    2.0
Name: NumberOfTrips, dtype: float64

Changing the Values

df.loc[df['NumberOfTrips'].isnull(), ['NumberOfTrips']] = 3

Checking the Changes

df.NumberOfTrips.isna().sum()

0

df.NumberOfTrips.describe()

count    4888.000000
mean        3.229746
std         1.822769
min         1.000000
25%         2.000000
50%         3.000000
75%         4.000000
max        22.000000
Name: NumberOfTrips, dtype: float64

sns.displot(df['NumberOfTrips'], kde = True)
plt.axvline(x = df.NumberOfTrips.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfTrips.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x249391926a0>

At the risk of sounding like a broken record, our data looks similar to how it was initially which we expected

Number of Children Visitng

df.NumberOfChildrenVisiting.isna().sum()

66

df.NumberOfChildrenVisiting.describe()

count    4822.000000
mean        1.187267
std         0.857861
min         0.000000
25%         1.000000
50%         1.000000
75%         2.000000
max         3.000000
Name: NumberOfChildrenVisiting, dtype: float64

sns.displot(df['NumberOfChildrenVisiting'], kde = True)
plt.axvline(x = df.NumberOfChildrenVisiting.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfChildrenVisiting.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x24939393970>

Observations

The median and mode are equal at one, and the mean is slightly higher at 1.18, we will impute the missing values with the median

Finding the Values

ReturnByPurchase('NumberOfChildrenVisiting')

Purchased Mean: 1.200438
Purchased Median: 1.0
Purchased Mode: 0    1.0
Name: NumberOfChildrenVisiting, dtype: float64

-----------

Unpurchased Mean: 1.18419
Unpurchased Median: 1.0
Unpurchased Mode: 0    1.0
Name: NumberOfChildrenVisiting, dtype: float64

Changing the Values

df.loc[df['NumberOfChildrenVisiting'].isnull(), ['NumberOfChildrenVisiting']] = 1

Checking the Changes

df.NumberOfChildrenVisiting.isna().sum()

0

df.NumberOfChildrenVisiting.describe()

count    4888.000000
mean        1.184738
std         0.852323
min         0.000000
25%         1.000000
50%         1.000000
75%         2.000000
max         3.000000
Name: NumberOfChildrenVisiting, dtype: float64

sns.displot(df['NumberOfChildrenVisiting'], kde = True)
plt.axvline(x = df.NumberOfChildrenVisiting.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfChildrenVisiting.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x2493924f580>

The median did not change, and the mean decreased by 0.003, so an insignificant change overall

Monthly Income

df.MonthlyIncome.isna().sum()

233

df.MonthlyIncome.describe()

count     4655.000000
mean     23619.853491
std       5380.698361
min       1000.000000
25%      20346.000000
50%      22347.000000
75%      25571.000000
max      98678.000000
Name: MonthlyIncome, dtype: float64

sns.displot(df['MonthlyIncome'], kde = True)
plt.axvline(x = df.MonthlyIncome.median(), color = 'black', ls='--')
plt.axvline(x = df.MonthlyIncome.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x2493951fb50>

Observations

We have a few pieces of data causing skewing to the right, with a massive max, as well as some outliers more than likely beneath our dataset. Out median and mean are fairly close in the grand scheme of things, so imputing with median will be helpful in bringing them closer.

Finding the Values

ReturnByPurchase('MonthlyIncome')

Purchased Mean: 22172.827703
Purchased Median: 21172.0
Purchased Mode: 0    17293.0
1    17404.0
2    20971.0
3    21082.0
Name: MonthlyIncome, dtype: float64

-----------

Unpurchased Mean: 23960.962835
Unpurchased Median: 22729.0
Unpurchased Mode: 0    20855.0
Name: MonthlyIncome, dtype: float64

Changing the Values

df.loc[(df['MonthlyIncome'].isnull() == True) & (df['ProdTaken'] == 1), ['MonthlyIncome']] = 21172
df.loc[(df['MonthlyIncome'].isnull() == True) & (df['ProdTaken'] == 0),['MonthlyIncome']] = 22729

Checking the Changes

df.MonthlyIncome.isna().sum()

0

df.MonthlyIncome.describe()

count     4888.000000
mean     23567.195376
std       5257.438805
min       1000.000000
25%      20485.000000
50%      22595.500000
75%      25424.750000
max      98678.000000
Name: MonthlyIncome, dtype: float64

sns.displot(df['NumberOfTrips'], kde = True)
plt.axvline(x = df.NumberOfTrips.median(), color = 'black', ls='--')
plt.axvline(x = df.NumberOfTrips.mean(), color = 'red')

<matplotlib.lines.Line2D at 0x2493a708070>

The mean dropped a little bit, and the median increased some, but the overall structure of the data stays mostly unchanged

Confirming no more Null Values

df.isna().sum()

ProdTaken                   0
Age                         0
TypeofContact               0
CityTier                    0
DurationOfPitch             0
Occupation                  0
Gender                      0
NumberOfPersonVisiting      0
NumberOfFollowups           0
ProductPitched              0
PreferredPropertyStar       0
MaritalStatus               0
NumberOfTrips               0
Passport                    0
PitchSatisfactionScore      0
OwnCar                      0
NumberOfChildrenVisiting    0
Designation                 0
MonthlyIncome               0
dtype: int64

Handling Outliers

df.describe().T

	count	mean	std	min	25%	50%	75%	max
ProdTaken	4888.0	0.188216	0.390925	0.0	0.0	0.0	0.00	1.0
Age	4888.0	37.557488	9.109545	18.0	31.0	37.0	43.00	61.0
CityTier	4888.0	1.654255	0.916583	1.0	1.0	1.0	3.00	3.0
DurationOfPitch	4888.0	15.381342	8.313127	5.0	9.0	13.0	19.00	127.0
NumberOfPersonVisiting	4888.0	2.905074	0.724891	1.0	2.0	3.0	3.00	5.0
NumberOfFollowups	4888.0	3.711129	0.998271	1.0	3.0	4.0	4.00	6.0
PreferredPropertyStar	4888.0	3.577946	0.797005	3.0	3.0	3.0	4.00	5.0
NumberOfTrips	4888.0	3.229746	1.822769	1.0	2.0	3.0	4.00	22.0
Passport	4888.0	0.290917	0.454232	0.0	0.0	0.0	1.00	1.0
PitchSatisfactionScore	4888.0	3.078151	1.365792	1.0	2.0	3.0	4.00	5.0
OwnCar	4888.0	0.620295	0.485363	0.0	0.0	1.0	1.00	1.0
NumberOfChildrenVisiting	4888.0	1.184738	0.852323	0.0	1.0	1.0	2.00	3.0
MonthlyIncome	4888.0	23567.195376	5257.438805	1000.0	20485.0	22595.5	25424.75	98678.0

Duration of Pitch

df.DurationOfPitch.describe()

count    4888.000000
mean       15.381342
std         8.313127
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max       127.000000
Name: DurationOfPitch, dtype: float64

histogramboxplot(data = df, feature = 'DurationOfPitch')

There are some pretty extreme outliers in Duration of Pitch, so cleaning up these outliers will really affect the way this data is structured, and hopefully produce better results. The ones close to the whisker are likely accurate, but the ones at 120 minutes seem like erronous entry, so I will set it to the median.

# df[df['DurationOfPitch'] > 120]

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
1434	0	37.0	Company Invited	3	126.0	Salaried	Male	2	3.0	Basic	3.0	Married	3.0	0	1	1	1.0	Executive	18482.0
3878	0	53.0	Company Invited	3	127.0	Salaried	Male	3	4.0	Basic	3.0	Married	4.0	0	1	1	2.0	Executive	22160.0

df.groupby("ProdTaken")["DurationOfPitch"].median()

ProdTaken
0    13.0
1    15.0
Name: DurationOfPitch, dtype: float64

df.loc[df['DurationOfPitch'] > 125, 'DurationOfPitch'] = 13

histogramboxplot(data = df, feature = 'DurationOfPitch')

This has improved the look of the dataset immediately, and makes it more approachable for further modification if needed, but ultimately likely will not

Number of Persons Visiting

df.DurationOfPitch.describe()

count    4888.000000
mean       15.334902
std         8.003437
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max        36.000000
Name: DurationOfPitch, dtype: float64

histogramboxplot(data = df, feature = 'NumberOfPersonVisiting')

This only has one outlier at the top, and it feels like it is likely an accurate value, so I will be leaving it for the time being

Number of Followups

df.DurationOfPitch.describe()

count    4888.000000
mean       15.334902
std         8.003437
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max        36.000000
Name: DurationOfPitch, dtype: float64

histogramboxplot(data = df, feature = 'NumberOfFollowups')

This one has a single datapoint above and a single one below that happens to be out of the whiskers, and seem accurate, so I will be leaving them be.

Number of Trips

df.DurationOfPitch.describe()

count    4888.000000
mean       15.334902
std         8.003437
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max        36.000000
Name: DurationOfPitch, dtype: float64

histogramboxplot(data = df, feature = 'NumberOfTrips')

This one was a debate, it is possible that the outliers at the top were accurate, but they also seem a bit extreme for the dataset, and may throw off a model. I have decided to take the values that are above 15 and impute them with the median, but leave the one near ten

# df[df['NumberOfTrips'] > 15]

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
385	1	30.0	Company Invited	1	10.0	Large Business	Male	2	3.0	Basic	3.0	Single	19.0	1	4	1	1.0	Executive	17285.0
816	0	39.0	Company Invited	1	15.0	Salaried	Male	3	3.0	Deluxe	4.0	Unmarried	21.0	0	2	1	0.0	Manager	21782.0
2829	1	31.0	Company Invited	1	11.0	Large Business	Male	3	4.0	Basic	3.0	Single	20.0	1	4	1	2.0	Executive	20963.0
3260	0	40.0	Company Invited	1	16.0	Salaried	Male	4	4.0	Deluxe	4.0	Unmarried	22.0	0	2	1	1.0	Manager	25460.0

df.groupby("ProdTaken")["NumberOfTrips"].median()

ProdTaken
0    3.0
1    3.0
Name: NumberOfTrips, dtype: float64

df.loc[df['NumberOfTrips'] > 18, 'NumberOfTrips'] = 3

histogramboxplot(data = df, feature = 'NumberOfTrips')

This seems to keep the shape of the data well intact, while leaving a likely accurate outlier in the data to prevent model overfit

Monthly Income

df.DurationOfPitch.describe()

count    4888.000000
mean       15.334902
std         8.003437
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max        36.000000
Name: DurationOfPitch, dtype: float64

histogramboxplot(data = df, feature = 'MonthlyIncome')

There are some outliers low, and a ton of outliers high. The ones below 40,000 seem likely accurate, but the ones approaching 100,000 a month seem likely erronous. I will examine the low values, determine if their data seems erronous, and then decide how to impute them. I will leave the values below 40,000 intact and impute the ones above 80,000 with the median

# df[df['MonthlyIncome'] < 10000]

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
142	0	38.0	Self Enquiry	1	9.0	Large Business	Female	2	3.0	Deluxe	3.0	Single	4.0	1	5	0	0.0	Manager	1000.0
2586	0	39.0	Self Enquiry	1	10.0	Large Business	Female	3	4.0	Deluxe	3.0	Single	5.0	1	5	0	1.0	Manager	4678.0

These values seem inaccurate given the fact that they are living in a major city, and are a manager in large busines. I will impute them with the median

# df[df['MonthlyIncome'] > 80000]

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
38	0	36.0	Self Enquiry	1	11.0	Salaried	Female	2	4.0	Basic	3.0	Divorced	1.0	1	2	1	0.0	Executive	95000.0
2482	0	37.0	Self Enquiry	1	12.0	Salaried	Female	3	5.0	Basic	5.0	Divorced	2.0	1	2	1	1.0	Executive	98678.0

df.groupby("ProdTaken")["MonthlyIncome"].median()

ProdTaken
0    22729.0
1    21172.0
Name: MonthlyIncome, dtype: float64

df.loc[df['MonthlyIncome'] > 80000, 'MonthlyIncome'] = 22729
df.loc[df['MonthlyIncome'] < 10000, 'MonthlyIncome'] = 22729

histogramboxplot(data = df, feature = 'MonthlyIncome')

This seems much better in terms of data, I will likely log scale this later due to the high number just outside the the upper whisker, but for now, I will leave it be.

Remove erronous Categorical Data

df['Gender'].value_counts()

Male       2916
Female     1817
Fe Male     155
Name: Gender, dtype: int64

df.loc[df['Gender'] == 'Fe Male', ['Gender']] = 'Female'

df['Gender'].value_counts()

Male      2916
Female    1972
Name: Gender, dtype: int64

Exploratory Data Analysis

Univariate Analysis

# df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3	3.0	Deluxe	3.0	Single	1.0	1	2	1	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	3	4.0	Deluxe	4.0	Divorced	2.0	0	3	1	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	2	3.0	Basic	3.0	Divorced	2.0	1	5	1	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	2	3.0	Basic	4.0	Divorced	1.0	0	5	1	0.0	Executive	18468.0

df.describe().T

	count	mean	std	min	25%	50%	75%	max
ProdTaken	4888.0	0.188216	0.390925	0.0	0.00	0.0	0.00	1.0
Age	4888.0	37.557488	9.109545	18.0	31.00	37.0	43.00	61.0
CityTier	4888.0	1.654255	0.916583	1.0	1.00	1.0	3.00	3.0
DurationOfPitch	4888.0	15.334902	8.003437	5.0	9.00	13.0	19.00	36.0
NumberOfPersonVisiting	4888.0	2.905074	0.724891	1.0	2.00	3.0	3.00	5.0
NumberOfFollowups	4888.0	3.711129	0.998271	1.0	3.00	4.0	4.00	6.0
PreferredPropertyStar	4888.0	3.577946	0.797005	3.0	3.00	3.0	4.00	5.0
NumberOfTrips	4888.0	3.215426	1.754188	1.0	2.00	3.0	4.00	8.0
Passport	4888.0	0.290917	0.454232	0.0	0.00	0.0	1.00	1.0
PitchSatisfactionScore	4888.0	3.078151	1.365792	1.0	2.00	3.0	4.00	5.0
OwnCar	4888.0	0.620295	0.485363	0.0	0.00	1.0	1.00	1.0
NumberOfChildrenVisiting	4888.0	1.184738	0.852323	0.0	1.00	1.0	2.00	3.0
MonthlyIncome	4888.0	23545.010434	5026.428314	16009.0	20486.75	22596.5	25407.75	38677.0

Age

descviolinbox(df, 'Age')

count    4888.000000
mean       37.557488
std         9.109545
min        18.000000
25%        31.000000
50%        37.000000
75%        43.000000
max        61.000000
Name: Age, dtype: float64

labeldisplot(df, 'Age')

bounds(df, 'Age')

Q1: 31.0
Q3: 43.0
IQR: 12.0
lower limit: 13.0
upper limit: 61.0

(13.0, 61.0, 31.0, 43.0)

numericdata(df, 'Age')

Mean: 37.557488
Median: 37.0
Mode: 37.0
Range: 12.0

Observations

Mean is 37.5 median is 37, mode is also 37 and the IQR is 12

• Median and mean are close, with the mean slightly to the right
• There is a slight right skew to the data, but not enough to fix with a log scale down the road
• We will probably want to scale our data, given that it has higher values than most of our data
• Interesting thing that might require looking at the dataset, some ages have double the count to those next to them, not something normally anticipated with age.

Duration of Pitch

descviolinbox(df, 'DurationOfPitch')

count    4888.000000
mean       15.334902
std         8.003437
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max        36.000000
Name: DurationOfPitch, dtype: float64

labeldisplot(df, 'DurationOfPitch')

bounds(df, 'DurationOfPitch')

Q1: 9.0
Q3: 19.0
IQR: 10.0
lower limit: -6.0
upper limit: 34.0

(-6.0, 34.0, 9.0, 19.0)

numericdata(df, 'DurationOfPitch')

Mean: 15.334902
Median: 13.0
Mode: 9.0
Range: 10.0

Observations

Mean is 15.3, Median is 13, Mode is 9, and the IQR is 10

• We have a pretty significant right skew in our data
• The KDE is important in this graph, because our initial thought was the Mode was 13
• Due to the long right tail, a log scale might be a good idea
• We also may want to just bin the data, as keeping this numeric might overfit the model

Monthly Income

descviolinbox(df, 'MonthlyIncome')

count     4888.000000
mean     23545.010434
std       5026.428314
min      16009.000000
25%      20486.750000
50%      22596.500000
75%      25407.750000
max      38677.000000
Name: MonthlyIncome, dtype: float64

labeldisplot(df, 'MonthlyIncome')

bounds(df, 'MonthlyIncome')

Q1: 20486.75
Q3: 25407.75
IQR: 4921.0
lower limit: 13105.25
upper limit: 32789.25

(13105.25, 32789.25, 20486.75, 25407.75)

numericdata(df, 'MonthlyIncome')

Mean: 23545.010434
Median: 22596.5
Mode: 22729.0
Range: 4921.0

Observations

Mean is \$23,545, Median is \$22,597, Mode is \$22,729 and the range is \$4,921

• This data has an extremely long right tale, and correspondingly a significant right skew
• The data is not very close to a normal distribution
• most of the right tail is flat, meaning there are roughly the same number of records
• There is a gap just below 20,000 which doesn't seem like accurate data, acquiring information for customers in this range may help us fit our model
• This seems to be a prime candidate for a log scale, as it has a long right tail, and the numeric values are significantly larger than the rest of the dataset

Prod Taken

labeledbarplot(df, 'ProdTaken')

Observations

Almost 4 times as many people did not purchase a product as those who did.

Type of Contact

labeledbarplot(df, 'TypeofContact')

Observations

Roughly double the number of people in our dataset contacted by self enquiry than Company Invited

City Tier

labeledbarplot(df, 'CityTier')

Observations

Most people do not live in a tier 2 city, and double the number of people live in Tier 1 cities than Tier 3

Occupation

labeledbarplot(df, 'Occupation')

Observations

• Most individuals are Salaried or Small Business
• Free Lancer has no significant bearing on our dataset
• Free Lancer and Large business should merge to prevent overfit, as an "Other" category

Gender

labeledbarplot(df, 'Gender', perc= True )

Observations

There are more men than women in this dataset, but it is much more representative than the data we normally see. There are roughly 1000 more men than women, or about a 60/40 split between men and women

Number of People Visiting

labeledbarplot(df, 'NumberOfPersonVisiting')

Observations

• This dataset is extremely close to a normal distribution
• 1 and 5 should be combined into an "other" category to prevent overfit, or 1+2 and 4+5 merge

Number of Followups

labeledbarplot(df, 'NumberOfFollowups')

Observations

• This again seems normally distributed
• Either 1, 2, and six should merge into "other", or more likely, one and two merge, and five and six merge.

Product Pitched

labeledbarplot(df, 'ProductPitched')

Observations

• This is not a normal distribution, with Basic and Deluxe topping the chart
• Basic, Deluxe, and Standard should remain the same while King and Super Deluxe merge into an "Other Category

Preferred Property Star

labeledbarplot(df, 'PreferredPropertyStar')

Observations

There are triple the number of people who prefer 3 star properties to 4 or 5

Marital Status

labeledbarplot(df, 'MaritalStatus')

Observations

• Married has double the number of records than other categories
• It is probbaly a good idea to simplify just to "Married" or "Unmarried"

Number of Trips

labeledbarplot(df, 'NumberOfTrips')

Observations

• This is not normally distributed, 2 and 3 have double the number of other records
• Merging categories for 6, 7, and 8 into "6 and Above" seems like a good idea to prevent overfit

Passport

labeledbarplot(df, 'Passport', perc = True)

Observations

70% of individuals in our dataset do not have a passport, which may significantly impact our sales of overseas packages

Pitch Satisfaction Score

labeledbarplot(df, 'PitchSatisfactionScore')

Observations

Our pitch satisfaction score records seem to be all fairly statistically significant and fairly normal in distribution. 3 seems to have the highest with roughly 500 records more than our other data. 2 is the lowest with a thousnad less records than 3

Car Ownership

labeledbarplot(df, 'OwnCar', perc =True)

Observations

Roughly 40% of people in our dataset did not own a car

Number of Children Visiting

labeledbarplot(df, 'NumberOfChildrenVisiting')

Observations

• Fairly normal distribution
• Combine 2 and 3 together to make 2+
• May be a good idea to merge with other datasets, such as NumberofPersonVisiting to create a simple category of if that group has children or not.

Designation

labeledbarplot(df, 'Designation')

Observations

• This is a fairly normal distribution
• I think there are too many categories for this, combining AVP, VP, and Senior Manager into just "other" makes sense, or AVP and VP can be combined into executive, and senior manager to manager

Bivariate Analysis

fig, ax = plt.subplots(figsize=(12,12))   
sns.heatmap(df.corr(), annot = True, cmap = "mako", square = True, ax = ax)

<AxesSubplot:>

Observations

• Most correlation with our dependent variable is extremely weak, and finding data to strengthen this is probably needed, maybe in categorical items that don't show up here
• Passport had the strongest correlation with the dependent variable at 0.26, but it did mean people with passports were more likely to purchase a package
• Age had a fairly strong correlation with monthly income at 0.46
• Number of Followups had a fairly weak correlation with number of people visiting
• Number of Person Visiting and Number of Children Visiting had a strong correlation
• The dependent variable had a 0.13 and 0.15 negative correlation with Monthly income and age respectively

sns.pairplot(df, diag_kind="kde", hue = 'ProdTaken')

<seaborn.axisgrid.PairGrid at 0x2493a99fd30>

Observations

Age and Duration of pitch have an extremely loose, or no correlation with monthy lincome

Numeric Means by ProdTaken

df.groupby(['ProdTaken']).mean().T

ProdTaken	0	1
Age	38.223286	34.685870
CityTier	1.615927	1.819565
DurationOfPitch	14.999496	16.781522
NumberOfPersonVisiting	2.901714	2.919565
NumberOfFollowups	3.657510	3.942391
PreferredPropertyStar	3.540071	3.741304
NumberOfTrips	3.203125	3.268478
Passport	0.233871	0.536957
PitchSatisfactionScore	3.044355	3.223913
OwnCar	0.622984	0.608696
NumberOfChildrenVisiting	1.181452	1.198913
MonthlyIncome	23871.228831	22138.016304

Observations

• People who purchase the packages are younger
• In general, people who purchased the package had a 17 minute pitch
• People who purchased a package had more followups than those who didn't
• People who purchased a package preferred higher property stars
• Passport holders were more likely to purchase a package
• Those more satisfied with the pitch were more likely to purchase a package
• Car owners were slightly less likely to purchase a package
• Those that made a lower income were more likely to purchase a package

Numeric Columns By ProdTaken

Age

stackedbox(df, 'Age', 'ProdTaken')

minmeanmax(df, 'Age', 'ProdTaken')

Not Purchased Min Mean and Max:

 Min: 18.0
 Mean: 38.223286
 Max: 61.0

Purchased Min Mean and Max:

 Min: 18.0
 Mean: 34.68587
 Max: 60.0

Observations

Those who purchased a package were slightly younger all around, with a lower median, mean, and Q3 point than those who didn't interestingly though they had the same minimum, but those who purchased had a lower maximum

Duration of Pitch

stackedbox(df, 'DurationOfPitch', 'ProdTaken')

minmeanmax(df, 'DurationOfPitch', 'ProdTaken')

Not Purchased Min Mean and Max:

 Min: 5.0
 Mean: 14.999496
 Max: 36.0

Purchased Min Mean and Max:

 Min: 6.0
 Mean: 16.781522
 Max: 36.0

Observations

• Those that purchased packages had a longer pitch in every way except for the maximum
• Those that didn't have some outliers due to having a lower mean, median, a smaller IQR, and a lower Q3 point
• If we saw significant correlation with duration of pitch it might be worth removing those errant values, but for now we will be leaving them

Monthly Income

stackedbox(df, 'MonthlyIncome', 'ProdTaken')

minmeanmax(df, 'MonthlyIncome', 'ProdTaken')

Not Purchased Min Mean and Max:

 Min: 16051.0
 Mean: 23871.228831
 Max: 38677.0

Purchased Min Mean and Max:

 Min: 16009.0
 Mean: 22138.016304
 Max: 38537.0

Observations

• We saw above that MonthlyIncome and Prodtaken had a slight negative correlation
• Given the values we see here, with a clear lower min, Q1, mean, median, Q3, and max the outliers are causing issues with that correlation
• I do not wish to remove or cap the outliers, as they are likely accurate, but doing a log scale probably helps us build a better model with the changes to the structure likely bringing in some of those outliers

Categorical Columns by ProdTaken

# df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3	3.0	Deluxe	3.0	Single	1.0	1	2	1	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	3	4.0	Deluxe	4.0	Divorced	2.0	0	3	1	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	2	3.0	Basic	3.0	Divorced	2.0	1	5	1	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	2	3.0	Basic	4.0	Divorced	1.0	0	5	1	0.0	Executive	18468.0

Type of Contact

stackedbarplot(df, 'TypeofContact', 'ProdTaken')

ProdTaken           0    1   All
TypeofContact                   
All              3968  920  4888
Self Enquiry     2859  610  3469
Company Invited  1109  310  1419
-------------------------------------------------------------------------------------------------------------------
ProdTaken               0         1
TypeofContact                      
Company Invited  0.781536  0.218464
Self Enquiry     0.824157  0.175843

sns.countplot(data= df, x='TypeofContact', hue = 'ProdTaken')

<AxesSubplot:xlabel='TypeofContact', ylabel='count'>

Observations

• Self Enquiries make up the majority of our data, and consequently share the most accepted products
• If you were invited by a company you were slightly more likely to purchase the product. with a little over 20% of customers purchasing the product

City Tier

stackedbarplot(df, 'CityTier', 'ProdTaken')

ProdTaken     0    1   All
CityTier                  
All        3968  920  4888
1          2670  520  3190
3          1146  354  1500
2           152   46   198
-------------------------------------------------------------------------------------------------------------------
ProdTaken         0         1
CityTier                     
3          0.764000  0.236000
2          0.767677  0.232323
1          0.836991  0.163009

sns.countplot(data= df, x='CityTier', hue = 'ProdTaken')

<AxesSubplot:xlabel='CityTier', ylabel='count'>

Obserations

• Tier 2 and 3 cities purchased products at a much higher rate than Tier 1 Cities
• Tier 1 Cities were offered many more packages than Tier 2 and 3 cities
• Tier 1 cities purchased the most packages, but Tier 3 is close with significantly fewer products offered
• tier 2 cities have very few records, so observations may be more skewed than Tier 1 or Tier 3 cities

Occupation

stackedbarplot(df, 'Occupation', 'ProdTaken')

ProdTaken          0    1   All
Occupation                     
All             3968  920  4888
Salaried        1954  414  2368
Small Business  1700  384  2084
Large Business   314  120   434
Free Lancer        0    2     2
-------------------------------------------------------------------------------------------------------------------
ProdTaken              0         1
Occupation                        
Free Lancer     0.000000  1.000000
Large Business  0.723502  0.276498
Small Business  0.815739  0.184261
Salaried        0.825169  0.174831

sns.countplot(data= df, x='Occupation', hue = 'ProdTaken')

<AxesSubplot:xlabel='Occupation', ylabel='count'>

df.loc[df['Occupation'] == 'Free Lancer']

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0	0.0	Executive	17090.0
2446	1	38.0	Self Enquiry	1	9.0	Free Lancer	Male	4	5.0	Basic	3.0	Single	8.0	1	3	0	1.0	Executive	20768.0

Observations

• Free Lancers take 100% of the packages they are offered, but there are exactly two records for freelancers.
• Removing the Free Lancer records and/or changing the label to one of the others will help us prevent disrupting the model
• Of the other models, large businesses took 30% of the packages offered, but had the smallest count of packages bought
• Small Business and Salaried both had a close purchase rate of about 20%, but salaried were offered more packages for roughly the same count

Gender

stackedbarplot(df, 'Gender', 'ProdTaken')

ProdTaken     0    1   All
Gender                    
All        3968  920  4888
Male       2338  578  2916
Female     1630  342  1972
-------------------------------------------------------------------------------------------------------------------
ProdTaken         0         1
Gender                       
Male       0.801783  0.198217
Female     0.826572  0.173428

sns.countplot(data= df, x='Gender', hue = 'ProdTaken')

<AxesSubplot:xlabel='Gender', ylabel='count'>

Observations

• Men purchased packages at a slightly higher rate than women
• Men were offered significantly more packages than women were
• Our resultant model may be slightly skewed to view men more favorably than women in package sales

Number of Persons Visiting

stackedbarplot(df, 'NumberOfPersonVisiting', 'ProdTaken')

ProdTaken                  0    1   All
NumberOfPersonVisiting                 
All                     3968  920  4888
3                       1942  460  2402
2                       1151  267  1418
4                        833  193  1026
1                         39    0    39
5                          3    0     3
-------------------------------------------------------------------------------------------------------------------
ProdTaken                      0         1
NumberOfPersonVisiting                    
3                       0.808493  0.191507
2                       0.811707  0.188293
4                       0.811891  0.188109
1                       1.000000  0.000000
5                       1.000000  0.000000

sns.countplot(data= df, x='NumberOfPersonVisiting', hue = 'ProdTaken')

<AxesSubplot:xlabel='NumberOfPersonVisiting', ylabel='count'>

Observations

• 1 and 5 people visiting did not purchase any packages
• 2,3, and 4 person groups all purchased about 20% of the packages they were offered
• There are 39 rows for 1 and 3 rows for 5, removing the rows outright is a bad idea
• Grouping 1 and 2, and 4 and 5 seems like a good idea, that does cause a slight negative skew to both, but overall they should stay roughly the same

Number of Followups

stackedbarplot(df, 'NumberOfFollowups', 'ProdTaken')

ProdTaken             0    1   All
NumberOfFollowups                 
All                3968  920  4888
4.0                1726  387  2113
3.0                1222  244  1466
5.0                 577  191   768
6.0                  82   54   136
2.0                 205   24   229
1.0                 156   20   176
-------------------------------------------------------------------------------------------------------------------
ProdTaken                 0         1
NumberOfFollowups                    
6.0                0.602941  0.397059
5.0                0.751302  0.248698
4.0                0.816848  0.183152
3.0                0.833561  0.166439
1.0                0.886364  0.113636
2.0                0.895197  0.104803

sns.countplot(data= df, x='NumberOfFollowups', hue = 'ProdTaken')

<AxesSubplot:xlabel='NumberOfFollowups', ylabel='count'>

Observations

• Six had the highest percentage of people purchase a package at roughly 40%
• One and two had the lowest percentage of takers with around 10% each
• There is a pretty significant correlation between number of followups and ProdTaken

Product Pitched

stackedbarplot(df, 'ProductPitched', 'ProdTaken')

ProdTaken          0    1   All
ProductPitched                 
All             3968  920  4888
Basic           1290  552  1842
Deluxe          1528  204  1732
Standard         618  124   742
King             210   20   230
Super Deluxe     322   20   342
-------------------------------------------------------------------------------------------------------------------
ProdTaken              0         1
ProductPitched                    
Basic           0.700326  0.299674
Standard        0.832884  0.167116
Deluxe          0.882217  0.117783
King            0.913043  0.086957
Super Deluxe    0.941520  0.058480

sns.countplot(data= df, x='ProductPitched', hue = 'ProdTaken')

<AxesSubplot:xlabel='ProductPitched', ylabel='count'>

Observations

• Basic Performed the Best in terms of sales with 30% of pitched products being sold
• The Super Deluxe performed the absolute worst, with just 6% of those that were pitched the product purchasing it
• The Deluxe was pitched a similar number of times to Basic, but performed far worse at just 10% of those having the product pitched purchasing the product
• There is a fairly strong negative correlation between products pitched and ProdTaken (assuming Super Deluxe or King are Higher numbers and Basic is the lowest)

Preferred Property Star

stackedbarplot(df, 'PreferredPropertyStar', 'ProdTaken')

ProdTaken                 0    1   All
PreferredPropertyStar                 
All                    3968  920  4888
3.0                    2531  488  3019
5.0                     706  250   956
4.0                     731  182   913
-------------------------------------------------------------------------------------------------------------------
ProdTaken                     0         1
PreferredPropertyStar                    
5.0                    0.738494  0.261506
4.0                    0.800657  0.199343
3.0                    0.838357  0.161643

sns.countplot(data= df, x='PreferredPropertyStar', hue = 'ProdTaken')

<AxesSubplot:xlabel='PreferredPropertyStar', ylabel='count'>

Observations

• 3 was preferred vastly more often over 4 and 5 star properties
• People who preferred 3, purchased a package only 16% of the time
• 4 and 5 were preferred by roughly an equal number of customers
• people who preferred 5 star properties purchased a packages 26% of the time, a massive leap of those that preferred 3 stars
• There is a positive correlation between property star and prodtaken

Marital Status

stackedbarplot(df, 'MaritalStatus', 'ProdTaken')

ProdTaken         0    1   All
MaritalStatus                 
All            3968  920  4888
Married        2014  326  2340
Single          612  304   916
Unmarried       516  166   682
Divorced        826  124   950
-------------------------------------------------------------------------------------------------------------------
ProdTaken             0         1
MaritalStatus                    
Single         0.668122  0.331878
Unmarried      0.756598  0.243402
Married        0.860684  0.139316
Divorced       0.869474  0.130526

sns.countplot(data= df, x='MaritalStatus', hue = 'ProdTaken')

<AxesSubplot:xlabel='MaritalStatus', ylabel='count'>

Observations

• Single people were by far the most likely to purchase a package, with 33% of individuals in that group purchasing a package
• Those that are married and divorced, the highest number of records in our dataset were the least likely to purchase a package at 14 and 13 percent each
• Single individuals have a similar count of package takers to married despite being vastly underrepresented in our dataset
• Unmarried individuals also had a fairly high percentage of takers with 24% buyinh a package
• These seems to be a strong correlation between those that have never married, and products being purchased

Number of Trips

stackedbarplot(df, 'NumberOfTrips', 'ProdTaken')

ProdTaken         0    1   All
NumberOfTrips                 
All            3968  920  4888
2.0            1165  299  1464
3.0             992  231  1223
1.0             508  112   620
6.0             258   64   322
5.0             396   62   458
7.0             156   62   218
4.0             417   61   478
8.0              76   29   105
-------------------------------------------------------------------------------------------------------------------
ProdTaken             0         1
NumberOfTrips                    
7.0            0.715596  0.284404
8.0            0.723810  0.276190
2.0            0.795765  0.204235
6.0            0.801242  0.198758
3.0            0.811120  0.188880
1.0            0.819355  0.180645
5.0            0.864629  0.135371
4.0            0.872385  0.127615

sns.countplot(data= df, x='NumberOfTrips', hue = 'ProdTaken')

<AxesSubplot:xlabel='NumberOfTrips', ylabel='count'>

Observations

• 7 and 8 top the list in terms of shares of packages sold, with roughly 28% each
• 1, 2, 3, and 6 sit in the middle around 20%
• 4 and 5 are the lowest with roughly 13% each
• There is a moderate correlation between number of trips and purchasing products
• Given the number of different categories, grouping 6 and above seems like a good idea, due to model overfit, however, the dataset is also currently numeric so this may be a bad idea

Passport

stackedbarplot(df, 'Passport', 'ProdTaken')

ProdTaken     0    1   All
Passport                  
All        3968  920  4888
1           928  494  1422
0          3040  426  3466
-------------------------------------------------------------------------------------------------------------------
ProdTaken         0         1
Passport                     
1          0.652602  0.347398
0          0.877092  0.122908

sns.countplot(data= df, x='Passport', hue = 'ProdTaken')

<AxesSubplot:xlabel='Passport', ylabel='count'>

Observation

• 35% of people who had a passport purchased a product vs 12% if they did not posses a passport
• There is a strong correlation between passport ownership and vacation taking, but passport owners are a smaller part of our dataset

Pitch Satisfaction Score

stackedbarplot(df, 'PitchSatisfactionScore', 'ProdTaken')

ProdTaken                  0    1   All
PitchSatisfactionScore                 
All                     3968  920  4888
3                       1162  316  1478
5                        760  210   970
4                        750  162   912
1                        798  144   942
2                        498   88   586
-------------------------------------------------------------------------------------------------------------------
ProdTaken                      0         1
PitchSatisfactionScore                    
5                       0.783505  0.216495
3                       0.786198  0.213802
4                       0.822368  0.177632
1                       0.847134  0.152866
2                       0.849829  0.150171

sns.countplot(data= df, x='PitchSatisfactionScore', hue = 'ProdTaken')

<AxesSubplot:xlabel='PitchSatisfactionScore', ylabel='count'>

Observations

• 3 and 5 top the list for packages purchased at roughly 20%, while 1 and 2 sit at the bottom with 15%
• 4 sits in the middle with 18% purchasing a package
• There is a weak correlation between pitch satisfaction and product purchase, but its hardly there

Own Car

stackedbarplot(df, 'OwnCar', 'ProdTaken')

ProdTaken     0    1   All
OwnCar                    
All        3968  920  4888
1          2472  560  3032
0          1496  360  1856
-------------------------------------------------------------------------------------------------------------------
ProdTaken         0         1
OwnCar                       
0          0.806034  0.193966
1          0.815303  0.184697

sns.countplot(data= df, x='OwnCar', hue = 'ProdTaken')

<AxesSubplot:xlabel='OwnCar', ylabel='count'>

Observations

• There appears to be almost no correlation between car ownership and product purchase
• We can probably drop this column in feature engineering, due to only a 0.9% difference between purchase rates

Number of Children Visiting

stackedbarplot(df, 'NumberOfChildrenVisiting', 'ProdTaken')

ProdTaken                    0    1   All
NumberOfChildrenVisiting                 
All                       3968  920  4888
1.0                       1747  399  2146
2.0                       1082  253  1335
0.0                        880  202  1082
3.0                        259   66   325
-------------------------------------------------------------------------------------------------------------------
ProdTaken                        0         1
NumberOfChildrenVisiting                    
3.0                       0.796923  0.203077
2.0                       0.810487  0.189513
0.0                       0.813309  0.186691
1.0                       0.814073  0.185927

sns.countplot(data= df, x='NumberOfChildrenVisiting', hue = 'ProdTaken')

<AxesSubplot:xlabel='NumberOfChildrenVisiting', ylabel='count'>

Observations

• This again had a very small variance in product purchase rates, only deviating between 18.6% and 20.3%, or a 1.7% difference
• We can likely drop this entire column to prevent model overfit

Designation

stackedbarplot(df, 'Designation', 'ProdTaken')

ProdTaken          0    1   All
Designation                    
All             3968  920  4888
Executive       1290  552  1842
Manager         1528  204  1732
Senior Manager   618  124   742
AVP              322   20   342
VP               210   20   230
-------------------------------------------------------------------------------------------------------------------
ProdTaken              0         1
Designation                       
Executive       0.700326  0.299674
Senior Manager  0.832884  0.167116
Manager         0.882217  0.117783
VP              0.913043  0.086957
AVP             0.941520  0.058480

sns.countplot(data= df, x='Designation', hue = 'ProdTaken')

<AxesSubplot:xlabel='Designation', ylabel='count'>

Observations

• Executives were far and away the most likely to purchase a package at 30%
• The next Highest was Senior Manager at 17%, almost half of the Executive numbers
• AVP was the lowest with just 5% purchasing a package
• Assuming the designation is supposed to be ordered Manager, Senior Manager, AVP, VP, Executive, there doesn't seem to be a strong correlation in the data, but there is something to senior managers and executives having a higher rate, that the data should be kept
• Grouping AVP and VP is a good idea

Multivariate Analysis of Different Features

Pointplot of Income, Number of Trips, and Product Taken

plt.figure(figsize=(15, 7))
sns.pointplot(x = 'NumberOfTrips', y='MonthlyIncome', data=df, hue = 'ProdTaken')

<AxesSubplot:xlabel='NumberOfTrips', ylabel='MonthlyIncome'>

Observations

The more you made, the less likely you were to buy any package, but the data seemed pretty similar in the variance of income by number of trips, with a major leap up from 2 trips a year to 3 in income

PointPlot of Income, Designation, and Product Taken

plt.figure(figsize=(15, 7))
sns.pointplot(x = 'Designation', y='MonthlyIncome', data=df, hue = 'ProdTaken')

<AxesSubplot:xlabel='Designation', ylabel='MonthlyIncome'>

Observations

• This absolutely blew us out of the water, we were expecting AVP and VP to be beneath executive in terms of pay, but executive makes much less than even manager.
• This weird pay gap could be explained by shares taken as income by executives rather than cash, but I am unfamiliar enough with the dataset that this is not a certainty
• Variance in pay for those that purchased a package is much higher than those that didn't, given the size of the error bars we see for AVP

PointPlot of Income, Product Piched, and ProdTaken

plt.figure(figsize=(15, 7))
sns.pointplot(x = 'ProductPitched', y='MonthlyIncome', data=df, hue = 'ProdTaken')

<AxesSubplot:xlabel='ProductPitched', ylabel='MonthlyIncome'>

Observations

• This graph is identical to the graph of pay above, meaning that individuals with the same designation were all pitched the same package with no variation
• This could have unintentional consequences for our dataset, and it may be a good idea to get more data than we currently have to account for this issue

PointPlot of Income, Preferred Property Star, and ProdTaken

plt.figure(figsize=(15, 7))
sns.pointplot(x = 'PreferredPropertyStar', y='MonthlyIncome', data=df, hue = 'ProdTaken')

<AxesSubplot:xlabel='PreferredPropertyStar', ylabel='MonthlyIncome'>

Observations

• those that preferred 3 star made more than those that preferred 4
• Five Star preference had a wide gap, with those preferring 5 star hotels and purchasing a package making less than 3 star preferrers who did the same
• People who made above 23000 a year did not purchase a package

PointPlot of Pitch Duration, Pitch Satisfaction, and Prod Taken

plt.figure(figsize=(15, 7))
sns.pointplot(x = 'PitchSatisfactionScore', y='DurationOfPitch', data=df, hue = 'ProdTaken')

<AxesSubplot:xlabel='PitchSatisfactionScore', ylabel='DurationOfPitch'>

Observations

• There is a lot of variance in pitch duration for those that purchased a product, with large error bars
• The longer the pitch was, the more likely people were to purchase a product
• Four star pitches tended to be shorter and less likely to result in a product being purchased than 3 or Five

Pointplot of Age, Marital Status, and ProdTaken

plt.figure(figsize=(15, 7))
sns.pointplot(data=df, x='MaritalStatus', y='Age', hue='ProdTaken')

<AxesSubplot:xlabel='MaritalStatus', ylabel='Age'>

Observations

• Single individuals who bought packages were the youngest people in our dataset
• Each category who purchased a product was younger than those who didn't
• Divorced and married individuals were the oldest in the dataset, and those that were divorced and bought a package had the widest spread of age in the dataset

Conclusions and Recommendations for EDA

Predictive Strength of Columns

Features that had strong predictors for our dependent variable were:

• Product Pitched
• Marital Status
• Number of Followups
• Passport
• Designation

These are important for determining whether or not someone purchases a package from our company. There were some more moderate predictors as well such as:

• Income
• Duration of Pitch
• City Tier
• Occupation
• Preferred Property Star
• Number of Trips

These weren't as important to the end result, but still has a bearing that is worth teasing out in our model. There are some weaker columns, that aren't as important such as:

• Age
• Type of Contact
• Pitch Satisfaction Score
• Gender

These were all extremely close value wise to one another, with little spread between values. they also, like pitch satisfaction score could have values that undermine the correlation the rest of the values had presented. They could be modified, or removed without significant impact to the model. Finally we have columns that appear to have no bearing on the data whatsoever:

• Number of Children
• Own car
• Number of Person visiting

These should all be removed to prevent overfit of the model.

NonNumeric Categories that Encoding

• Type of Contact
• Occupation
• Gender
• Product Pitched
• Marital Status
• Designation

Feature Engineering

Remove Unnecessary Columns

Own Car

This had no bearing on whether or not an individual purchased a package, and should be removed due to potential model overfitting

stackedbarplot(df, 'OwnCar', 'ProdTaken')

ProdTaken     0    1   All
OwnCar                    
All        3968  920  4888
1          2472  560  3032
0          1496  360  1856
-------------------------------------------------------------------------------------------------------------------
ProdTaken         0         1
OwnCar                       
0          0.806034  0.193966
1          0.815303  0.184697

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	OwnCar	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3	3.0	Deluxe	3.0	Single	1.0	1	2	1	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	3	4.0	Deluxe	4.0	Divorced	2.0	0	3	1	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	2	3.0	Basic	3.0	Divorced	2.0	1	5	1	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	2	3.0	Basic	4.0	Divorced	1.0	0	5	1	0.0	Executive	18468.0

df.drop('OwnCar', axis = 1, inplace = True)

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3	3.0	Deluxe	3.0	Single	1.0	1	2	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	3	4.0	Deluxe	4.0	Divorced	2.0	0	3	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	2	3.0	Basic	3.0	Divorced	2.0	1	5	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	2	3.0	Basic	4.0	Divorced	1.0	0	5	0.0	Executive	18468.0

Number of Person Visiting

This too, had no bearing on whether or not a vaction package was purchased for 2, 3, and 4. one and five each did not purchase a package, but there were only 44 total records among the two categories. I am removing it to prevent model overfit on such a minor piece of data

stackedbarplot(df, 'NumberOfPersonVisiting', 'ProdTaken')

ProdTaken                  0    1   All
NumberOfPersonVisiting                 
All                     3968  920  4888
3                       1942  460  2402
2                       1151  267  1418
4                        833  193  1026
1                         39    0    39
5                          3    0     3
-------------------------------------------------------------------------------------------------------------------
ProdTaken                      0         1
NumberOfPersonVisiting                    
3                       0.808493  0.191507
2                       0.811707  0.188293
4                       0.811891  0.188109
1                       1.000000  0.000000
5                       1.000000  0.000000

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfPersonVisiting	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3	3.0	Deluxe	3.0	Single	1.0	1	2	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	3	4.0	Deluxe	4.0	Divorced	2.0	0	3	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	3	4.0	Basic	3.0	Single	7.0	1	3	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	2	3.0	Basic	3.0	Divorced	2.0	1	5	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	2	3.0	Basic	4.0	Divorced	1.0	0	5	0.0	Executive	18468.0

df.drop('NumberOfPersonVisiting', axis = 1, inplace = True)

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3.0	Deluxe	3.0	Single	1.0	1	2	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	4.0	Deluxe	4.0	Divorced	2.0	0	3	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	4.0	Basic	3.0	Single	7.0	1	3	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	3.0	Basic	3.0	Divorced	2.0	1	5	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	3.0	Basic	4.0	Divorced	1.0	0	5	0.0	Executive	18468.0

Number of Children Visiting

This is the third column that showed no bearing on whether or not a product was purchased, and again, will be removed to prevent model overfit.

stackedbarplot(df, 'NumberOfChildrenVisiting', 'ProdTaken')

ProdTaken                    0    1   All
NumberOfChildrenVisiting                 
All                       3968  920  4888
1.0                       1747  399  2146
2.0                       1082  253  1335
0.0                        880  202  1082
3.0                        259   66   325
-------------------------------------------------------------------------------------------------------------------
ProdTaken                        0         1
NumberOfChildrenVisiting                    
3.0                       0.796923  0.203077
2.0                       0.810487  0.189513
0.0                       0.813309  0.186691
1.0                       0.814073  0.185927

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	NumberOfChildrenVisiting	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3.0	Deluxe	3.0	Single	1.0	1	2	0.0	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	4.0	Deluxe	4.0	Divorced	2.0	0	3	2.0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	4.0	Basic	3.0	Single	7.0	1	3	0.0	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	3.0	Basic	3.0	Divorced	2.0	1	5	1.0	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	3.0	Basic	4.0	Divorced	1.0	0	5	0.0	Executive	18468.0

df.drop('NumberOfChildrenVisiting', axis = 1, inplace = True)

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3.0	Deluxe	3.0	Single	1.0	1	2	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	4.0	Deluxe	4.0	Divorced	2.0	0	3	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	4.0	Basic	3.0	Single	7.0	1	3	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	3.0	Basic	3.0	Divorced	2.0	1	5	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	3.0	Basic	4.0	Divorced	1.0	0	5	Executive	18468.0

Pitch Satisfaction Score

We have two columns, pitch duration and pitch satisfaction score that describe an individual pitch. While Pitch Satisfaction score could be useful, 3 and 5 both have a larger number of takers than 4, and Pitch Duration tends to show a stronger correlation with ProdTaken than Pitch Satisfaction does. This may negatively affect a model down the road, but we do not have the background to know for sure.

stackedbarplot(df, 'PitchSatisfactionScore', 'ProdTaken')

ProdTaken                  0    1   All
PitchSatisfactionScore                 
All                     3968  920  4888
3                       1162  316  1478
5                        760  210   970
4                        750  162   912
1                        798  144   942
2                        498   88   586
-------------------------------------------------------------------------------------------------------------------
ProdTaken                      0         1
PitchSatisfactionScore                    
5                       0.783505  0.216495
3                       0.786198  0.213802
4                       0.822368  0.177632
1                       0.847134  0.152866
2                       0.849829  0.150171

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	PitchSatisfactionScore	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3.0	Deluxe	3.0	Single	1.0	1	2	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	4.0	Deluxe	4.0	Divorced	2.0	0	3	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	4.0	Basic	3.0	Single	7.0	1	3	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	3.0	Basic	3.0	Divorced	2.0	1	5	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	3.0	Basic	4.0	Divorced	1.0	0	5	Executive	18468.0

df.drop('PitchSatisfactionScore', axis = 1, inplace = True)

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3.0	Deluxe	3.0	Single	1.0	1	Manager	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	4.0	Deluxe	4.0	Divorced	2.0	0	Manager	20130.0
2	1	37.0	Self Enquiry	1	8.0	Free Lancer	Male	4.0	Basic	3.0	Single	7.0	1	Executive	17090.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	3.0	Basic	3.0	Divorced	2.0	1	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	3.0	Basic	4.0	Divorced	1.0	0	Executive	18468.0

Clean up and Consolidate labels in Categorical featues

Occupation

sns.countplot(data= df, x='Occupation', hue = 'ProdTaken')

<AxesSubplot:xlabel='Occupation', ylabel='count'>

Freelancer is a statistically insignificant amount of data, and so we decided to drop it. It could negatively affect our data, and we wanted to ensure that our model didn't assume every Freelancer was going to purchase a package

df.drop(df[(df['Occupation'] == 'Free Lancer')].index, inplace=True)

sns.countplot(data= df, x='Occupation', hue = 'ProdTaken')

<AxesSubplot:xlabel='Occupation', ylabel='count'>

Number of Trips

stackedbarplot(df, 'NumberOfTrips', 'ProdTaken')

ProdTaken         0    1   All
NumberOfTrips                 
All            3968  918  4886
2.0            1165  299  1464
3.0             992  231  1223
1.0             508  112   620
6.0             258   64   322
5.0             396   62   458
4.0             417   61   478
7.0             156   61   217
8.0              76   28   104
-------------------------------------------------------------------------------------------------------------------
ProdTaken             0         1
NumberOfTrips                    
7.0            0.718894  0.281106
8.0            0.730769  0.269231
2.0            0.795765  0.204235
6.0            0.801242  0.198758
3.0            0.811120  0.188880
1.0            0.819355  0.180645
5.0            0.864629  0.135371
4.0            0.872385  0.127615

We have eight different categories for trips, and some of them share similar probabilities for package purchase rates. If we leave them be, we may overfit a model, so binning them (before later reconverting them to numeric fields) should help us prevent this. Due to how they are arranged with purchase rates we would like to group them like the following

• 1-3
• 4-5
• 6
• 7+

This will halve the number of catefories in our dataset and keep the general preditictive strengths of each one. 6 as a value is difficult due to not fitting well with 4 and five or seven and eight, and better fitting with 1-3. We just left it on its own to not make the eventual data difficult to understand

sns.countplot(data= df, x='NumberOfTrips', hue = 'ProdTaken')

<AxesSubplot:xlabel='NumberOfTrips', ylabel='count'>

conditions = [
    (df['NumberOfTrips'] >= 7),
    (df['NumberOfTrips'] == 6),
    (df['NumberOfTrips'] >= 4) & (df['NumberOfTrips'] <= 5),
    (df['NumberOfTrips'] >= 1) & (df['NumberOfTrips'] <= 3)
]

values = [
    '7+',
    '6',
    '4-5',
    '1-3'
]

df['NumberOfTrips'] = np.select(conditions, values)

sns.countplot(data=df, x='NumberOfTrips', order=['1-3', '4-5', '6', '7+'], hue='ProdTaken')

<AxesSubplot:xlabel='NumberOfTrips', ylabel='count'>

stackedbarplot(df, 'NumberOfTrips', 'ProdTaken')

ProdTaken         0    1   All
NumberOfTrips                 
All            3968  918  4886
1-3            2665  642  3307
4-5             813  123   936
7+              232   89   321
6               258   64   322
-------------------------------------------------------------------------------------------------------------------
ProdTaken             0         1
NumberOfTrips                    
7+             0.722741  0.277259
6              0.801242  0.198758
1-3            0.805866  0.194134
4-5            0.868590  0.131410

We can see that there was no significant chantges to the % of buyers in our data by grouping the numbers this way, so our model should be able to predict well without being overfit

Designation

stackedbarplot(df, 'Designation', 'ProdTaken')

ProdTaken          0    1   All
Designation                    
All             3968  918  4886
Executive       1290  550  1840
Manager         1528  204  1732
Senior Manager   618  124   742
AVP              322   20   342
VP               210   20   230
-------------------------------------------------------------------------------------------------------------------
ProdTaken              0         1
Designation                       
Executive       0.701087  0.298913
Senior Manager  0.832884  0.167116
Manager         0.882217  0.117783
VP              0.913043  0.086957
AVP             0.941520  0.058480

sns.countplot(data=df, x='Designation', hue='ProdTaken')

<AxesSubplot:xlabel='Designation', ylabel='count'>

plt.figure(figsize=(15, 7))
sns.pointplot(data=df, x='Designation', y='MonthlyIncome')

<AxesSubplot:xlabel='Designation', ylabel='MonthlyIncome'>

This data was a little bit weird both in its initial labelling and how we hope to consolidate it. Executives made the least, followed by managers, then senior managers, AVPs, and finally VPs made the most. We want to group VP and AVP, and then group Manager and Senior manager while leaving executive alone.

conditions = [
    (df['Designation'] == 'Manager'), 
    (df['Designation'] == 'Executive'),
    (df['Designation'] == 'Senior Manager'),
    (df['Designation'] == 'AVP'),
    (df['Designation'] == 'VP')
]

values = [
    'Management',
    'Executive',
    'Management',
    'VP and AVP',
    'VP and AVP'
]

df['Designation'] = np.select(conditions, values)

stackedbarplot(df, 'Designation', 'ProdTaken')

ProdTaken       0    1   All
Designation                 
All          3968  918  4886
Executive    1290  550  1840
Management   2146  328  2474
VP and AVP    532   40   572
-------------------------------------------------------------------------------------------------------------------
ProdTaken           0         1
Designation                    
Executive    0.701087  0.298913
Management   0.867421  0.132579
VP and AVP   0.930070  0.069930

sns.countplot(data=df, x='Designation', hue='ProdTaken')

<AxesSubplot:xlabel='Designation', ylabel='count'>

The rough idea of the percentage of buyers has stayed intact after we have reduced the number of columns in Designation, and grouped them together. This also more accurately reflects the number of individuals set to be an executive

Scale Numeric Columns

df.head(5) 

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	41.0	Self Enquiry	3	6.0	Salaried	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	20993.0
1	0	49.0	Company Invited	1	14.0	Salaried	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	20130.0
3	0	33.0	Company Invited	1	9.0	Salaried	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	17909.0
4	0	37.0	Self Enquiry	1	8.0	Small Business	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	18468.0
5	0	32.0	Company Invited	1	8.0	Salaried	Male	3.0	Basic	3.0	Single	1-3	0	Executive	18068.0

We have three numeric columns that have a wider range than the rest of our dataset that will need to be scaled accordingly. These are Age, Duration of Pitch and Monthly Income. Age has relatively few outliers, so it will just be min-maxed scaled to preserve the rough shape of the data. Monthly Income and Duration of Pitch have a long right tail, and will be log scaled to change the datashape towards a normal distribution

Age

labeldisplot(df, 'Age')

We can see our data is close to normal, and wouldn't benefit greatly from log scaling

scaler.fit(df[['Age']])
df[['Age']] = scaler.transform(df[['Age']])

labeldisplot(df, 'Age')

We can see the shape of our data remains intact, with its slight right skew but otherwise mostly normal distribution

Duration of Pitch

labeldisplot(df, 'DurationOfPitch')

sns.kdeplot(data=df, x="DurationOfPitch", fill=False)

<AxesSubplot:xlabel='DurationOfPitch', ylabel='Density'>

df.DurationOfPitch.describe()

count    4886.000000
mean       15.337700
std         8.003873
min         5.000000
25%         9.000000
50%        13.000000
75%        19.000000
max        36.000000
Name: DurationOfPitch, dtype: float64

We don't want the data to stay this way, and a log scale will help us reign in this long right tail, that is all within the whiskers of the data, shifting the center to the right

df['DurationOfPitch'] = np.log(df['DurationOfPitch'])

df.DurationOfPitch.describe()

count    4886.000000
mean        2.603731
std         0.499056
min         1.609438
25%         2.197225
50%         2.564949
75%         2.944439
max         3.583519
Name: DurationOfPitch, dtype: float64

sns.kdeplot(data=df, x="DurationOfPitch", fill=False)

<AxesSubplot:xlabel='DurationOfPitch', ylabel='Density'>

labeldisplot(df, 'DurationOfPitch')

This data is still not normal, but it is significantly closer than it would be otherwise, and should help prevent skewing our model, but this requires further testing

Monthly Income

labeldisplot(df, 'MonthlyIncome')

This data has the vague structure of a normal distribution, but it has a long right tail that we want to remove and make the dataset as a whole more normal.

sns.kdeplot(data=df, x="MonthlyIncome", fill=False)

<AxesSubplot:xlabel='MonthlyIncome', ylabel='Density'>

df.MonthlyIncome.describe()

count     4886.000000
mean     23546.899918
std       5026.451397
min      16009.000000
25%      20487.250000
50%      22597.000000
75%      25411.250000
max      38677.000000
Name: MonthlyIncome, dtype: float64

df['MonthlyIncome'] = np.log(df['MonthlyIncome'])

df.MonthlyIncome.describe()

count    4886.000000
mean       10.045997
std         0.199718
min         9.680906
25%         9.927558
50%        10.025572
75%        10.142947
max        10.563000
Name: MonthlyIncome, dtype: float64

sns.kdeplot(data=df, x="MonthlyIncome", fill=False)

<AxesSubplot:xlabel='MonthlyIncome', ylabel='Density'>

labeldisplot(df, 'MonthlyIncome')

This at least shortened the tail, but didn't do an insane amount to normalize the distribution. We will now need to go through and do a min/max encoding due to the numbers being larger than we want, especially in comparison to the rest of the data

scaler.fit(df[['MonthlyIncome']])
df[['MonthlyIncome']] = scaler.transform(df[['MonthlyIncome']])

labeldisplot(df, 'MonthlyIncome')

We have kept the shape of the data, but left it between 0 and 1.

Encoding Categorical Values

df.head()

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	Self Enquiry	3	1.791759	Salaried	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	Company Invited	1	2.639057	Salaried	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	Company Invited	1	2.197225	Salaried	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	Self Enquiry	1	2.079442	Small Business	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	Company Invited	1	2.079442	Salaried	Male	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

Type of Contact, Occupation, gender, ProductPitched, Marital Status, Number of Trips, and Designation are all non-numeric and will need to be encoded using either onehot or label encoding. Given that we already have a large number of columns, we will be utilizing Lable Encoding for these values.

Type of Contact

labeledbarplot(df, 'TypeofContact')

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	Self Enquiry	3	1.791759	Salaried	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	Company Invited	1	2.639057	Salaried	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	Company Invited	1	2.197225	Salaried	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	Self Enquiry	1	2.079442	Small Business	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	Company Invited	1	2.079442	Salaried	Male	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

Transforming the Data with Label Encoder

df['TypeofContact'] = labelEncoder.fit_transform(df['TypeofContact'])

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	Salaried	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	Salaried	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	Salaried	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	Small Business	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	Salaried	Male	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

labeledbarplot(df, 'TypeofContact')

Occupation

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	Salaried	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	Salaried	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	Salaried	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	Small Business	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	Salaried	Male	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

labeledbarplot(df, 'Occupation')

Transforming the Data with Label Encoder

df['Occupation'] = labelEncoder.fit_transform(df['Occupation'])

labeledbarplot(df, 'Occupation')

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	Male	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

Gender

labeledbarplot(df, 'Gender')

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	Female	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	Male	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	Female	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	Male	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	Male	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

Transforming the Data with Label Encoder

df['Gender'] = labelEncoder.fit_transform(df['Gender'])

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

labeledbarplot(df, 'Gender')

Product Pitched

labeledbarplot(df, 'ProductPitched')

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	Deluxe	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	Deluxe	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	Basic	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	Basic	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	Basic	3.0	Single	1-3	0	Executive	0.137164

Transforming the Data with Label Encoder

df['ProductPitched'] = labelEncoder.fit_transform(df['ProductPitched'])

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	Single	1-3	0	Executive	0.137164

labeledbarplot(df, 'ProductPitched')

Marital Status

labeledbarplot(df, 'MaritalStatus')

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	Single	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	Divorced	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	Divorced	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	Divorced	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	Single	1-3	0	Executive	0.137164

Transforming the Data with Label Encoder

df['MaritalStatus'] = labelEncoder.fit_transform(df['MaritalStatus'])

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	2	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	0	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	0	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	0	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	2	1-3	0	Executive	0.137164

labeledbarplot(df, 'MaritalStatus')

Number of Trips

labeledbarplot(df, 'NumberOfTrips')

df.head(10)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	2	1-3	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	0	1-3	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	0	1-3	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	0	1-3	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	2	1-3	0	Executive	0.137164
6	0	0.953488	1	1	2.197225	2	0	2.0	0	5.0	0	4-5	1	Executive	0.111912
7	0	0.279070	1	1	3.401197	1	1	3.0	0	3.0	1	1-3	0	Executive	0.113387
8	0	0.465116	0	1	3.367296	1	1	4.0	3	3.0	3	1-3	0	Management	0.483602
9	0	0.418605	1	1	3.496508	2	1	3.0	1	3.0	0	7+	0	Management	0.265688
10	0	0.395349	1	1	3.091042	2	1	2.0	0	4.0	0	1-3	0	Executive	0.096149

Transforming the Data with Label Encoder

df['NumberOfTrips'] = labelEncoder.fit_transform(df['NumberOfTrips'])

df.head(10)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	2	0	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	0	0	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	0	0	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	0	0	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	2	0	0	Executive	0.137164
6	0	0.953488	1	1	2.197225	2	0	2.0	0	5.0	0	1	1	Executive	0.111912
7	0	0.279070	1	1	3.401197	1	1	3.0	0	3.0	1	0	0	Executive	0.113387
8	0	0.465116	0	1	3.367296	1	1	4.0	3	3.0	3	0	0	Management	0.483602
9	0	0.418605	1	1	3.496508	2	1	3.0	1	3.0	0	3	0	Management	0.265688
10	0	0.395349	1	1	3.091042	2	1	2.0	0	4.0	0	0	0	Executive	0.096149

labeledbarplot(df, 'NumberOfTrips')

Designation

labeledbarplot(df, 'Designation')

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	2	0	1	Management	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	0	0	0	Management	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	0	0	1	Executive	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	0	0	0	Executive	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	2	0	0	Executive	0.137164

Transforming the Data with Label Encoder

df['Designation'] = labelEncoder.fit_transform(df['Designation'])

df.head(5)

	ProdTaken	Age	TypeofContact	CityTier	DurationOfPitch	Occupation	Gender	NumberOfFollowups	ProductPitched	PreferredPropertyStar	MaritalStatus	NumberOfTrips	Passport	Designation	MonthlyIncome
0	1	0.534884	1	3	1.791759	1	0	3.0	1	3.0	2	0	1	1	0.307267
1	0	0.720930	0	1	2.639057	1	1	4.0	1	4.0	0	0	0	1	0.259678
3	0	0.348837	0	1	2.197225	1	0	3.0	0	3.0	0	0	1	0	0.127143
4	0	0.441860	1	1	2.079442	2	1	3.0	0	4.0	0	0	0	0	0.161988
5	0	0.325581	0	1	2.079442	1	1	3.0	0	3.0	2	0	0	0	0.137164

labeledbarplot(df, 'Designation')

Features that Remain Untouched

We had a few features that we didn't touch in our Feature Engineering and we wanted to walk through what we had yet to cover in our engineering.

• Prodtaken was our dependent variable, and had already been encoded to be between 0 and 1 in the dataset, it didn't need any more engineering from us.

• City Tier was alread numeric between 1 and 3. We could have done label encoding, but we felt it may be confusing, and the range was within features that had already been encoded in a similar way

• Number of followups had six different categories, one for each number of followup. We considered grouping and then encoding them again, but there was good variance in prodtaken for each column, and so decided to leave it be.

• Preferred property star was a debate for us, we considered doing label encoding, since the array for it started at 3 rather than 0, but all the ranges were within those covered by number of followups, and we didn't know how it would affect the model one way or the other due to lack of knowledge about how models are made

• Passport was similar to ProdTaken, with it already having been encoded and given a value between 0 and 1, and as such didn't need touching by us