Building and Understanding a Classification Model for Customer Loan Approval thru PyCaret

How to develop a customer loan approval classification model and easily create iterations using PyCaret

Dummy Bank noticed that it has an increasing volume of bad loans and wants to act on it by building a Machine Learning model based on its historical data of “good” and “bad” loans. Additionally, the company asks to find the factors driving these bad loans so the business can take the necessary actions moving forward.

Importing the libraries

Loading the data

Initial EDA + Initial Data Pre-Processing

Data needs to undergo iterations of pre-processing and EDA. As such, the following shall be implemented:

Autoviz

Data Transformation

After the data has completed the pre-processing and EDA stage, the following data transformation will be performed:

Pycaret Stratified Sampling of training and test set

Classification Model Customization

In order to achieve better metrics, the following customizations were implemented:

Model Building

Pycaret XGB:

Pycaret Random Forest:

Understanding the Key Drivers thru SHAP

With all the steps conducted, the following can be concluded:

2) Baseline model (XGB) is performing with Recall of 63%. This was achieved by adjusting class weights and tweaking the classification threshold to 10%.

3) In line with these findings, the following are recommended:

Related posts:

It does not matter what work we do , rather how we do it , is more supreme . One person is submitting assignment after the deadline and other one before deadline. In this case , work is same but how…

Most of us earn money in the form of a monthly salary. Then we spend it in buying necessities and luxuries. If there is anything left over, we save or invest it in some ‘safe’ avenues, take out a…

Well I am down 3 Pounds I am watching what I eat all things trying to stay as healthy as I know still watching calories and using smaller portions, I am trying to stay under 1500 calories a day. I…