Predicting Default Risk - Creditworthiness

Classification Modelling and Performance Comparison

by Sooyeon Won

Keywords

• Analytical Framework
• Supervised Learning
• Binary Classification Models
  • Logistic Regression Model
  • Decision Tree Model
  • Random Forest Model
  • Gradient Boosting Model
  • AdaBoost Model
• ROC curve
• K-fold cross-validation
• Model Selection and its Application

Summary of Findings

A bank recently received an influx of loan applications. I built and apply different classification models to provide an appropriate recommendation. Among the various models, I found out that Random Forest Model is the most suitable for predicting the creditworthiness of credit applicants. Based on the Random Forest model, I concluded 408 customers belong to the segment 'Creditworthy’.

This analysis is separated into three parts.

1. The first part of the analysis (Filename: 01_Creditworthiness_Data_Cleaning_Exploration.ipynb) contains the process of Data Cleaning and Data Exploration with appropriate visualizations. At the end of the Part1, I select the features associated with the creditworthiness of customers, which is the target variable of the analysis.
2. In the second part of the analysis (Filename: 02_Creditworthiness_Data_Analysis.ipynb), I trained the dataset with various classification models: Logistic Regression, Decision Tree, Random Forest, Gradient Boosted Model, and AdaBoost Model. Then they are validated with the test dataset. Additionally, I evaluate the performance of the models with k-fold cross validation techniques.
3. After selecting the model with best performance, I conclude that how many new credit applicants are classified as creditworthy customers.

References

Logistic Regression (aka logit, MaxEnt) classifier - sklearn
How to Calculate Feature Importance With Python
Decision Tree Classifier - sklearn
Categorical Features and Encoding in Decision Trees