🏦 Loan Default Prediction & Risk Strategy

📌 Project Overview

Financial institutions incur significant losses when borrowers default on loans. This project analyzes historical loan data to identify patterns associated with high-risk borrowers. By building a predictive machine learning model, the project aims to automate the loan eligibility process and suggest specific risk mitigation strategies to reduce the bank's non-performing assets (NPA).

📂 Project Files

• Loan Default Prediction & Risk Strategy.ipynb: The technical Jupyter Notebook containing data cleaning, EDA, feature engineering, and model training.
• Loan Default Prediction & Risk Strategy.pdf: A comprehensive report summarizing the findings, visualizations, and business recommendations.

🛠️ Tech Stack

• Language: Python 3.x
• Data Processing: Pandas, NumPy
• Visualization: Matplotlib, Seaborn
• Machine Learning: Scikit-Learn (Logistic Regression, Random Forest, Decision Trees)
• Metrics: Precision, Recall, F1-Score, ROC-AUC (Focus on minimizing False Negatives)

📊 Key Workflow

1. Data Understanding & Cleaning

• Handled missing values in demographic and financial columns.
• Detected and treated outliers in Income and Loan Amount.
• Analyzed class imbalance (Defaults vs. Non-Defaults).

2. Exploratory Data Analysis (EDA)

• Financial Health: Analyzed Debt-to-Income (DTI) ratios and their correlation with default rates.
• Demographics: Examined the impact of employment length, home ownership, and education level on repayment behavior.
• Loan Characteristics: Investigated interest rates and loan grades assigned by the bank.

3. Machine Learning Modeling

Different classification models were tested to predict the binary outcome (Default / No Default):

• Logistic Regression: Used as a baseline for interpretability.
• Random Forest Classifier: Implemented to capture non-linear relationships and feature importance.
• Model Optimization: Tuned hyperparameters to improve Recall (Sensitivity), as missing a defaulter is more costly than flagging a good borrower.

💡 Risk Strategy & Business Insights

Based on the analysis, the following strategies are recommended:

1. Strict DTI Thresholds: Applicants with a Debt-to-Income ratio above a certain threshold (e.g., 40%) show a significantly higher probability of default.
2. Interest Rate Adjustment: High-risk profiles identified by the model should be offered loans at adjusted interest rates to cover potential losses (Risk-Based Pricing).
3. Enhanced Verification: employment verification should be mandatory for applicants requesting high loan amounts with short credit histories.

🚀 How to Run

1. Clone the repository:

   git clone:  https://github.com/amit95ranjan/Loan-Default-Prediction-Risk-Strategy

2. Install dependencies:

   pip install pandas numpy matplotlib seaborn scikit-learn

3. Run the Notebook:

   jupyter notebook "Loan Default Prediction & Risk Strategy.ipynb"

---

Created by Amit Ranjan - Data Analyst