🏦 Loan Approval Prediction System

A data engineering and machine learning project that predicts loan approval decisions based on applicant financial and personal information — built with Python, Scikit-learn, and Flask/Streamlit.

---

📌 Overview

This project covers the complete data pipeline — from raw JSON ingestion and EDA to model training, serialization, and a deployed prediction app. It demonstrates real-world ETL practices, feature engineering, and ML model deployment.

---

📁 Project Structure

Data-Engineering-Project-2/
│
├── EDA_Train_pickle.py       # Exploratory data analysis + model training + pickle export
├── app.py                    # Web app for loan approval prediction
├── loan_approval_model.pkl   # Trained ML model (serialized)
│
├── applicant_info.json       # Applicant personal data
├── financial_info.json       # Applicant financial data
├── loan_info.json            # Loan request details
│
├── test_load_data.py         # Unit tests for data loading
├── requirements.txt          # Project dependencies
└── README.md

---

⚙️ How It Works

JSON Data Files  →  EDA & Cleaning  →  Feature Engineering  →  Model Training  →  Pickle Export  →  Web App

1. Ingest — Load applicant, financial, and loan data from JSON files
2. EDA — Explore distributions, handle missing values, encode categories
3. Train — Train a classification model (Scikit-learn) to predict approval
4. Serialize — Export trained model as loan_approval_model.pkl
5. Serve — Load model in app.py and predict on new applicant data

---

🛠️ Tech Stack

Layer	Tools
Language	Python
Data handling	Pandas, NumPy
ML Model	Scikit-learn
Serialization	Pickle
App	Flask / Streamlit
Testing	Pytest
Data format	JSON

---

🚀 Getting Started

1. Clone the repo

git clone https://github.com/SahilUjgare/Data-Engineering-Project-2.git
cd Data-Engineering-Project-2

2. Install dependencies

pip install -r requirements.txt

3. Train the model

python EDA_Train_pickle.py

4. Run the app

python app.py

5. Run tests

python test_load_data.py

---

📊 Model Details

• Task — Binary classification (Loan Approved / Rejected)
• Input features — Applicant info, financial history, loan amount, tenure
• Output — Approval prediction with confidence score
• Model file — loan_approval_model.pkl

---

📂 Data Files

File	Description
applicant_info.json	Name, age, employment, credit history
financial_info.json	Income, expenses, existing loans
loan_info.json	Requested amount, tenure, purpose

---

📄 License

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