Delinquency Model

By: Deepak Mittal

Analysis Date: November 01, 2020




What is delinquency?

It is a condition when a person is unable to do an expected activity at its scheduled time. Our study relates to the prediction of loan transactions whether a customer will pay back the loan on its scheduled time or not. It also includes how different parameters affect delinquency.



Heat Map

Heat map shows the relation between features and its magnitude using the column bar.

Heat Map


Data Analysis

No doubt, delinquency depends upon the various factor, some of these factors are discussed below.


1. Daily amount spent

Daily amount spent

It is less risky to loan someone whose 30-day average of daily amount spent from main account is greater than 40000.


2. Main account balance(30 days)

Main account balance(30 days)

It is less risky to loan someone whose average main account balance over last 30 days is greater than 50000.


3. Count main account recharge(30 days)

Count main account recharge(30 days)

It is less risky to loan someone who got recharged main account for more than 25 times in the last 30 days.


4. Sum main account recharge(30 days)

Sum main account recharge(30 days)

It is less risky to loan someone whose total amount of recharge in the main account over last 30 days is greater than 60000.


5. Count main account recharge(90 days)

Count main account recharge(90 days)

It is less risky to loan someone whose main account got recharged for more than 50 times in the last 90 days.


6. Sum main account recharge(90 days)

Sum main account recharge(90 days)

It is less risky to loan someone whose total amount of recharge in the main account over last 90 days is greater than 100000.


7. Count loans(30 days)

Count loans(30 days)

It is less risky to loan someone who had taken loan for more than 20 times in the last 30 days.


8. Total amount(30 days)

Total amount(30 days)f

It is less risky to loan someone whose total amount of loan in the last 30 days is greater than 100.



Machine Learning Modeling

1. ExtraTreesClassifier

ExtraTreesClassifier Distribution ExtraTreesClassifier Features importance


2. RandomForestClassifier

RandomForestClassifier Distribution RandomForestClassifier Features importance


3. GradientBoostingClassifier

GradientBoostingClassifier Distribution GradientBoostingClassifier Features importance


4. LogisticsRegression

LogisticsRegression Distribution


5. DecisionTreeClassifier

DecisionTreeClassifier Distribution DecisionTreeClassifier Features importance


6. SupportVectorClassifier

SupportVectorClassifier Distribution


7. KNeighborsClassifier

KNeighborsClassifier Distribution



Model scores

Models






Webpage Maintained & Designed by Deepak Mittal | Complete Code

Data Source: algo8.ai