1 Imports

2 Data Collection

## Rows: 381,109
## Columns: 12
## $ id                   <dbl> 1, 2, 3, 4, 5, 6, 7…
## $ Gender               <chr> "Male", "Male", "Ma…
## $ Age                  <dbl> 44, 76, 47, 21, 29,…
## $ Driving_License      <dbl> 1, 1, 1, 1, 1, 1, 1…
## $ Region_Code          <dbl> 28, 3, 28, 11, 41, …
## $ Previously_Insured   <dbl> 0, 0, 0, 1, 1, 0, 0…
## $ Vehicle_Age          <chr> "> 2 Years", "1-2 Y…
## $ Vehicle_Damage       <chr> "Yes", "No", "Yes",…
## $ Annual_Premium       <dbl> 40454, 33536, 38294…
## $ Policy_Sales_Channel <dbl> 26, 26, 26, 152, 15…
## $ Vintage              <dbl> 217, 183, 27, 203, …
## $ Response             <dbl> 1, 0, 1, 0, 0, 0, 0…

3 Data Cleaning

## Rows: 381,109
## Columns: 12
## $ id                   <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15…
## $ gender               <fct> male, male, male, male, female, female, male, fem…
## $ age                  <dbl> 44, 76, 47, 21, 29, 24, 23, 56, 24, 32, 47, 24, 4…
## $ driving_license      <fct> yes, yes, yes, yes, yes, yes, yes, yes, yes, yes,…
## $ region_code          <dbl> 28, 3, 28, 11, 41, 33, 11, 28, 3, 6, 35, 50, 15, …
## $ previously_insured   <fct> no, no, no, yes, yes, no, no, no, yes, yes, no, y…
## $ vehicle_age          <fct> over_2_years, between_1_2_years, over_2_years, be…
## $ vehicle_damage       <fct> yes, no, yes, no, no, yes, yes, yes, no, no, yes,…
## $ health_annual_paid   <dbl> 40454, 33536, 38294, 28619, 27496, 2630, 23367, 3…
## $ policy_sales_channel <dbl> 26, 26, 26, 152, 152, 160, 152, 26, 152, 152, 124…
## $ days_associated      <dbl> 217, 183, 27, 203, 39, 176, 249, 72, 28, 80, 46, …
## $ response             <fct> yes, no, yes, no, no, no, no, yes, no, no, yes, n…
Characteristic N = 381,1091
gender
    female 175,020 (45.92%)
    male 206,089 (54.08%)
age 36 (25, 49)
driving_license
    yes 380,297 (99.79%)
    no 812 (0.21%)
region_code 28 (15, 35)
previously_insured
    yes 174,628 (45.82%)
    no 206,481 (54.18%)
vehicle_age
    below_1_year 164,786 (43.24%)
    between_1_2_years 200,316 (52.56%)
    over_2_years 16,007 (4.20%)
vehicle_damage
    yes 192,413 (50.49%)
    no 188,696 (49.51%)
health_annual_paid 31,669 (24,405, 39,400)
policy_sales_channel 133 (29, 152)
days_associated 154 (82, 227)
response
    yes 46,710 (12.26%)
    no 334,399 (87.74%)
1 n (%); Median (IQR)

3.1 Data Types

4 Column Description

variables description
id Unique ID for the customer
gender Gender of the customer
age Age of the customer
driving_license Customer has DL (yes/no)
region_code Unique code for the region of the customer
previously_insured Customer already has Vehicle Insurance (yes/no)
vehicle_age Age of the Vehicle
vehicle_damage Customer got his/her vehicle damaged in the past (yes/no)
health_annual_paid The amount customer needs to pay as premium in the year
policy_sales_channel Anonymized Code for the channel of outreaching to the customer ie. Different Agents, Over Mail, Over Phone, In Person, etc.
days_associated Number of Days, Customer has been associated with the company
response Customer is interested in car insurance (yes/no)

5 Descriptive Statistics

## Rows: 381,109
## Columns: 12
## $ id                   <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15…
## $ gender               <fct> male, male, male, male, female, female, male, fem…
## $ age                  <dbl> 44, 76, 47, 21, 29, 24, 23, 56, 24, 32, 47, 24, 4…
## $ driving_license      <fct> yes, yes, yes, yes, yes, yes, yes, yes, yes, yes,…
## $ region_code          <dbl> 28, 3, 28, 11, 41, 33, 11, 28, 3, 6, 35, 50, 15, …
## $ previously_insured   <fct> no, no, no, yes, yes, no, no, no, yes, yes, no, y…
## $ vehicle_age          <fct> over_2_years, between_1_2_years, over_2_years, be…
## $ vehicle_damage       <fct> yes, no, yes, no, no, yes, yes, yes, no, no, yes,…
## $ health_annual_paid   <dbl> 40454, 33536, 38294, 28619, 27496, 2630, 23367, 3…
## $ policy_sales_channel <dbl> 26, 26, 26, 152, 152, 160, 152, 26, 152, 152, 124…
## $ days_associated      <dbl> 217, 183, 27, 203, 39, 176, 249, 72, 28, 80, 46, …
## $ response             <fct> yes, no, yes, no, no, no, no, yes, no, no, yes, n…

5.1 Check data structure

Data summary
Name df_cleaned
Number of rows 381109
Number of columns 12
_______________________
Column type frequency:
factor 6
numeric 6
________________________
Group variables None

Variable type: factor

skim_variable n_missing complete_rate ordered n_unique top_counts
gender 0 1 FALSE 2 mal: 206089, fem: 175020
driving_license 0 1 FALSE 2 yes: 380297, no: 812
previously_insured 0 1 FALSE 2 no: 206481, yes: 174628
vehicle_age 0 1 FALSE 3 bet: 200316, bel: 164786, ove: 16007
vehicle_damage 0 1 FALSE 2 yes: 192413, no: 188696
response 0 1 FALSE 2 no: 334399, yes: 46710

Variable type: numeric

skim_variable n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
id 0 1 190555.00 110016.84 1 95278 190555 285832 381109 ▇▇▇▇▇
age 0 1 38.82 15.51 20 25 36 49 85 ▇▃▃▂▁
region_code 0 1 26.39 13.23 0 15 28 35 52 ▃▂▇▃▃
health_annual_paid 0 1 30564.39 17213.16 2630 24405 31669 39400 540165 ▇▁▁▁▁
policy_sales_channel 0 1 112.03 54.20 1 29 133 152 163 ▅▁▁▃▇
days_associated 0 1 154.35 83.67 10 82 154 227 299 ▇▇▇▇▇

5.2 Numerical Statistics

age days_associated health_annual_paid
Mean 38.82 154.35 30564.39
Std.Dev 15.51 83.67 17213.16
Min 20.00 10.00 2630.00
Q1 25.00 82.00 24405.00
Median 36.00 154.00 31669.00
Q3 49.00 227.00 39400.00
Max 85.00 299.00 540165.00
MAD 17.79 108.23 11125.43
IQR 24.00 145.00 14995.00
CV 0.40 0.54 0.56
Skewness 0.67 0.00 1.77
SE.Skewness 0.00 0.00 0.00
Kurtosis -0.57 -1.20 34.00
N.Valid 381109.00 381109.00 381109.00
Pct.Valid 100.00 100.00 100.00

6 Visualization

6.1 Numerical Attributes

6.2 Categorical Attributes

7 Hypothesis Validation

7.0.1 H1) Older customers are more likely to be interested in car insurance.

Characteristic yes, N = 46,7101 no, N = 334,3991 p-value2
age 43 (35, 51) 34 (24, 49) <0.001
1 Median (IQR)
2 Wilcoxon rank sum test

7.0.1.1 Young people seems to be less likely interested in car insurance. The median age for interested customers is 43 years (IQR: 35, 51), while the median for non-interested customers is 34 years (IQR: 24, 49). ## explain the p-value

7.0.2 H2) Women are likely more interested in car insurance 🟥

Characteristic yes, N = 46,7101 no, N = 334,3991
gender
    female 18,185 (39%) 156,835 (47%)
    male 28,525 (61%) 177,564 (53%)
1 n (%)

7.0.2.1 For customers interested in car insurance, 61% were men, and 39% were women. So this hypothesis is FALSE. Although, gender and response are statistically significant, i. e., are related.

7.0.3 H3) Customers having newer cars are more likely to be interested in car insurance

Characteristic yes, N = 46,7101 no, N = 334,3991 p-value2
vehicle_age <0.001
    below_1_year 7,202 (15%) 157,584 (47%)
    between_1_2_years 34,806 (75%) 165,510 (49%)
    over_2_years 4,702 (10%) 11,305 (3.4%)
1 n (%)
2 Pearson’s Chi-squared test

7.0.3.1 Customers who own a car between 1 and 2 years are more likely to be interested in the car insurance (75%). While, only 15% of the interested customers have new car.

7.0.4 H4) Customer with previous car damage are more likely to accept car insurance ✅

Characteristic yes, N = 46,7101 no, N = 334,3991 p-value2
vehicle_damage 45,728 (98%) 146,685 (44%) <0.001
1 n (%)
2 Pearson’s Chi-squared test

7.0.4.1 Customers with previous car damage are more likely to be interested in car insurance, as 98% said yes.

7.0.5 H5) Customers with previous car insurance are more likely to accept car insurance :

Characteristic yes, N = 46,7101 no, N = 334,3991 p-value2
previously_insured 158 (0.3%) 174,470 (52%) <0.001
1 n (%)
2 Pearson’s Chi-squared test

7.0.5.1 Only 0.3& of customers interested in car insurance have car previously insured

7.0.6 H6) Interest in car insurance is greater in customers which have higher annual health insurance 🟥

Characteristic yes, N = 46,7101 no, N = 334,3991 p-value2
health_annual_paid 33,002 (24,868, 41,297) 31,504 (24,351, 39,120) <0.001
1 Median (IQR)
2 Wilcoxon rank sum test

Although the health annual paid showed to be significant, we consider this hyphotesis false and will further investigate the outliers, in he next cycle for example.

7.0.7 H7) Customers who has health insurance for LONGER are more likely to be interested in car insurance 🟥

Characteristic yes, N = 46,7101 no, N = 334,3991 p-value2
days_associated 154 (82, 226) 154 (82, 227) 0.5
1 Median (IQR)
2 Wilcoxon rank sum test

7.0.7.1 This Hypothesis is false, basically the interested customers and non-interested customers have the same amount of days associated. Yes(median:154 days, IQR: 82, 226); No (154 days, IQR: 82, 227).

7.0.8 Hypothesis Conclusion

hypothesis conclusion relevance
H1) OLDER customers are more likely to be interested in car insurance True High
H2) Women are likely more interested in car insurance False Medium
H3) Customers having newer cars are more likely to be interested in car insurance False High
H4) Customer with previous car damage are more likely to accept car insurance True High
H5) Customers with previous car insurance are more likely to accept car insurance False High
H6) Interest in car insurance is greater in customers which have higher annual health insurance False Low
H7) Customers who has health insurance for LONGER are more likely to be interested in car insurance False Low

8 Multivariable Analysis

8.0.1 Correlation Matrix

8.0.2 Visualization

There is no high correlation between numerical variables

9 Data Preparation

Frequency encoding for policy_sales_channel

Target encoding for gender e region_code

9.1 Target encoding

9.2 Frequency encoding

9.2.1 Using the created encoders in the dataset

9.3 Splitting into train and test datasets

9.3.1 Check response proportions

Characteristic N = 285,8311
response
    yes 35,032 (12%)
    no 250,799 (88%)
1 n (%)
Characteristic N = 95,2781
response
    yes 11,678 (12%)
    no 83,600 (88%)
1 n (%)

9.3.2 Using tidymodels steps to continue the preprocessing

9.3.3 Applying the recipe

10 Feature Selection

## Time difference of 7.408854 secs

10.0.1 Show results

In tgis first cycle we are going to select the seven most importatnt vaiables according to the mean decrease gini

## [1] "vehicle_damage_no"     "days_associated"       "age"                  
## [4] "health_annual_paid"    "previously_insured_no" "policy_sales_channel" 
## [7] "region_code"