Home Credit Score Card Model

Created by Fitria Dwi Wulandari – July, 2022

Project Background

Many individuals struggle to obtain loans due to insufficient or non-existent credit histories. Home Credit aims to enhance financial inclusion for the unbanked population by offering a positive and secure borrowing experience. To ensure this underserved group has access to loans, Home Credit leverages alternative data to assess clients' repayment abilities.

Objectives

This project aims to reduce the rejection of creditworthy clients by:

• Identifying characteristics of clients who face difficulties in repaying loans.
• Developing a predictive model to accurately assess clients' repayment abilities.

Methodology

Data Preparation

• Source: Data obtained from the Home Credit Data Scientist Virtual Internship Program at Rakamin Academy.
• Actions: Cleaning and preparing the data for analysis to ensure high-quality inputs for the model.

Exploratory Data Analysis (EDA)

• Purpose: To discover patterns, spot anomalies, and gain a deeper understanding of the data's characteristics.
• Techniques: Visualization, summary statistics, and correlation analysis to identify significant features.

Machine Learning

• Goal: Build models to predict clients' repayment capabilities.
• Approach: 6 algorithms were evaluated to determine the best prediction model.

Tools

• Programming Language: Python
• Libraries: pandas, numpy, scikit-learn, matplotlib, seaborn

Results

The analysis revealed that the Random Forest model, with an accuracy of 99%, was the most effective in predicting clients' repayment capabilities. This model enables precise and reliable decision-making, ensuring that creditworthy clients are approved and provided with suitable loan terms, ultimately empowering them for successful repayment and financial stability.

Future Work

• Model Improvement: Explore more advanced machine learning algorithms and feature engineering techniques to improve model accuracy.
• Broader Application: Extend the analysis to other aspects of loan approval, such as loan terms and customer segmentation, to further enhance the approval process.

Repository Contents

• Scripts: Python scripts for data preprocessing, EDA, and model building.
• Report Deck: Detailed reports on findings and model performance.