🧠 Employee Attrition Predictor – Streamlit App

This interactive web app helps HR professionals and analysts predict whether an employee is at risk of leaving (attrition) based on their profile. The app uses a machine learning model trained with XGBoost and explained using SHAP (SHapley Additive Explanations).

---

🚀 Features

• ✅ Predict employee attrition with a tuned XGBoost model
• 🎯 Uses a custom probability threshold for realistic classification
• 📊 Interactive SHAP explainability for transparent predictions
• 🧮 Real-time risk scoring with adjustable profile inputs
• 💡 Clean and simple UI built with Streamlit

---

🛠 How to Run Locally

1. Clone this repo

git clone https://github.com/your-username/employee-attrition-app.git
cd employee-attrition-app

2. Install dependencies

pip install -r requirements.txt

3. Launch the app

streamlit run app.py

---

🌐 Live Demo

👉 Click to try the app

---

📁 File Structure

📦 employee-attrition-app/
├── app.py                      # Streamlit frontend
├── employee.py                 # Model training and export
├── xgb_employee_model.pkl      # Trained XGBoost model
├── final_model_features.pkl    # Feature names used by the model
├── final_default_values.pkl    # Default values used for prediction UI
├── dropdown_options.pkl        # Categorical dropdown values
├── requirements.txt            # Dependencies
└── README.md                   # You're reading this!

---

📊 Model Details

• Algorithm: XGBoost Classifier
• Tuning: RandomizedSearchCV with ROC AUC scoring
• Handling Imbalance: scale_pos_weight
• Explainability: SHAP TreeExplainer for bar/force plots

---

🤝 Contributors

👤 Amey Suresh Borkar
📧 amey.borkar01@gmail.com
🔗 LinkedIn | GitHub

---

📄 License

MIT License — use it, share it, build on it.