/Credit-Risk-Prediction

Primary LanguageJupyter Notebook

Credit Risk Prediction Model for Loan Eligibility Assessment

Project Goals

  • Develop a model to predict loan approval category (P1-P4 representing risk levels) for new customers.
  • Leverage the model to aid loan approval decisions.

Data

  • Over 50,000 customer records with features including demographics, financial information, and loan history.
  • Target variable: Categorical variable representing loan approval category (P1-P4).

Methodology

Data Cleaning

  • Identified and addressed missing values using appropriate techniques.
  • Handled data inconsistencies.
  • Performed exploratory data analysis (EDA) to understand data distribution and relationships.

Feature Engineering

  • Conducted Chi-square tests to identify significant relationships between categorical features and the target variable.
  • Used VIF analysis to remove highly correlated numerical features.
  • Implemented domain knowledge to create new features if necessary.

Model Building and Evaluation

  • Built an XGBoost model to predict loan approval category.
  • Tuned hyperparameters to optimize model performance.
  • Evaluated model on unseen data using accuracy metrics.

Results

  • Achieved 85% accuracy on unseen data.
  • Developed a model effective in assessing credit risk and aiding loan approval decisions.

Getting Started

This repository includes:

  • Jupyter notebooks for data cleaning, feature engineering, model building, and evaluation.