LOVE_PREDICT_ML_MODEL

What It Does

This machine learning model predicts relationship compatibility based on behavioral, emotional, and psychological features from two individuals. The system classifies their connection into one of three categories:

• 0 – Just Friends
• 1 – Could Work Out
• 2 – Strong Match

It uses a trained XGBoost classifier (94.5% accuracy) and supports Random Forest (88.6% accuracy) as an alternative. A web-based quiz interface collects answers for 30 relationship-based questions and sends them to the ML backend for prediction.

---

Setup & Installation

git clone https://github.com/<your-username>/LOVE_PREDICT_ML_MODEL.git
cd LOVE_PREDICT_ML_MODEL

# (Optional) Create virtual environment
python3 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Or manually:

pip install pandas scikit-learn xgboost joblib flask

---

Training the Model

python train_model.py

This will:

• Load your dataset (CSV)
• Train the XGBoost model
• Save:
  • xgb_model.pkl – trained model
  • scaler.pkl – fitted StandardScaler

---

How to Use

• Start your Flask server:

python app.py

• Send a POST request with all 30 features:

curl -X POST http://localhost:5000/predict \
     -H "Content-Type: application/json" \
     -d '{ "communication_frequency": 8, "trust_score": 9, ..., "reaction_to_praise": 2 }'

• Output:

{
  "prediction": 2
}

---

Web UI

The frontend (HTML/CSS/JS) provides:

• A landing page with explanation
• A 30-question quiz (1 question at a time)
• Disclaimer + accuracy notice
• Sends input to /predict endpoint and shows compatibility result

---

Folder Structure

LOVE_PREDICT_ML_MODEL/
│
├── model/
│   ├── xgb_model.pkl          # Trained model
│   ├── scaler.pkl             # StandardScaler
│
├── backend/
│   ├── app.py                 # Flask app for prediction
│   ├── train_model.py         # Model training script
│
├── frontend/
│   ├── index.html             # Landing page
│   ├── quiz.html              # Quiz form
│   ├── styles.css             # UI styling
│   ├── script.js              # Question logic + animation
│
├── data/
│   ├── your_dataset.csv       # 30-feature CSV with labels
│
└── README.md

---

Disclaimer

• This tool is for entertainment and educational use only.
• It is not meant for serious relationship decisions.
• Accuracy:
  • XGBoost: 94.5%
  • Random Forest: 88.6%

---

Dependencies

• pandas
• scikit-learn
• xgboost
• joblib
• flask
• (Optional) uvicorn, fastapi if using a FastAPI backend

---

Author

Deepan
Made with for AI + Relationship Fun

---

License

This project is open-sourced under the MIT License.