Credit Risk Prediction Model

Created by Fitria Dwi Wulandari – August, 2022

Project Background

XYZ Company operates in the lending industry. In the lending industry, making decisions about loan approvals is crucial. When a lending company receives a loan application, it must assess the applicant's profile to determine if they are likely to repay the loan. Approving a loan to an applicant who is unlikely to repay it can lead to significant financial losses for the company. Conversely, rejecting a loan application from an applicant who is likely to repay it results in missed business opportunities. Therefore, identifying and minimizing bad-risk loans is essential to mitigate financial losses and maintain business profitability.

Objectives

This project aims to minimize the approval of bad-risk loans by:

• Identifying patterns that indicate a high likelihood of loan default (bad risk).
• Developing predictive models to assess the credit risk of loan applicants.

Methodology

Data Preparation

• Source: Data obtained from the ID/X Partners Data Scientist Virtual Internship Program at Rakamin Academy.
• Actions: Cleaning and preparing the data for analysis, ensuring data quality and consistency.

Exploratory Data Analysis (EDA)

• Purpose: Discover patterns, spot anomalies, and gain a deeper understanding of the data's characteristics.

Machine Learning

• Approach: Building predictive models to assess credit risk.
• Algorithms Tested: Eight different algorithms were evaluated to determine the best model for credit risk prediction.

Tools

• Programming Language: Python.
• Libraries: Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn.

Results

• The analysis identified key factors influencing loan repayment likelihood: last payment month, last payment amount, principal amount received, recovery value, and last payment year.
• The Random Forest model, with an accuracy of 99%, was the most effective in predicting credit risk. This model enables precise and reliable decision-making, significantly reducing the likelihood of granting loans to bad-risk applicants and thereby minimizing financial losses.

Future Work

• Model Improvement: Further hypertuning the models and exploring advanced algorithms to improve performance.
• Extended Analysis: Include additional data sources and variables for a more comprehensive understanding of credit risk factors.

Repository Contents

• Script: Python scripts for data preprocessing, cleaning, and model training.
• Report Deck: Detailed reports on findings, including insights and model performance metrics.