pranavvyawahare25/AI_Predictive_Model_for_Credit_Underwriting

AIPredictiveModelforCredit_Underwriting

This repository features an ML-powered predictive model for automating loan eligibility assessments. It analyzes demographic, financial, and credit data to predict loan approvals with high accuracy. Using advanced algorithms, it enhances decision-making, reduces manual effort, and ensures fair, transparent credit evaluation.

📊 Credit Underwriting - Loan Approval Prediction

📌 Overview

This project is a machine learning-based credit underwriting system that predicts whether a loan will be approved or defaulted. It uses an XGBoost model trained on a credit risk dataset containing financial and personal details of applicants.

🚀 Features

• —Predict loan approval status based on applicant details.
• —Handles missing data using median imputation.
• —Encodes categorical features for model compatibility.
• —Feature scaling using StandardScaler.
• —Hyperparameter tuning using GridSearchCV.
• —Streamlit web application for easy user interaction.

🛠️ Technologies Used

• —Python 🐍
• —Pandas, NumPy
• —Scikit-learn
• —XGBoost
• —Matplotlib, Seaborn
• —Streamlit (for UI)

📂 File Structure

📁 Credit_Underwriting
│── AI_Credit_Status.py    # Streamlit web app
│── train_model.py         # Model training script
│── credit_model.pkl       # Trained XGBoost model
│── label_encoders.pkl     # Label encoders for categorical features
│── scaler.pkl             # StandardScaler for normalization
│── requirements.txt       # Dependencies list
│── README.md              # Project documentation

📊 Dataset

• —Target Variable: loan_status (Approved/Defaulted)
• —Features Used:
• —person_emp_length: Employment length in years
• —loan_int_rate: Interest rate on the loan
• —person_home_ownership: Type of home ownership (encoded)
• —loan_intent: Purpose of the loan (encoded)
• —loan_grade: Loan grade assigned to the applicant (encoded)
• —cb_person_default_on_file: History of default (encoded)
• —annual_income: Annual income of the applicant
• —credit_score: Credit score of the applicant
• —loan_amount: Loan amount requested
• —current_debt: Existing debt of the applicant
• —previous_defaults: Number of past loan defaults

🎯 Model Training

• —Uses XGBClassifier (XGBoost) for high accuracy.
• —Hyperparameter tuning with GridSearchCV.
• —Evaluation metrics:
• —Accuracy Score
• —ROC-AUC Score
• —Classification Report (Precision, Recall, F1-score)
• —Saves trained model, encoders, and scaler for deployment.

🖥️ Running the Project

1️⃣ Install Dependencies

pip install -r requirements.txt

2️⃣ Train the Model

python train_model.py

3️⃣ Run the Streamlit App

streamlit run AI_Credit_Status.py

📜 License

This project is open-source and available under the MIT License.

👨‍💻 Author

Developed by Saloni Waghmare 🚀

Feel free to ⭐ the repository if you find this project useful! 😊 http://localhost:8501 http://192.168.43.220:8501