RetroML

Welcome to RetroML, a nostalgic, config-driven machine learning pipeline that brings the radical vibes of the '90s to modern AutoML workflows. Run interpretable ML models through JSON configs or interactively via UI demos — no coding required!

Quick Start

1. Install Dependencies

   pip install -r requirements.txt

2. Run a Pipeline

   python retroml.py configs/classification_example.json

3. Try the Interactive UI (if you have the UI file)

   python retroml_ui.py

Run the Demo for RetroML Classification Demo

Demo Focus:

The included classification demo specifically demonstrates:

• Customer Churn Prediction (Binary Classification)
  • Predicts whether customers will leave (churn=1) or stay (churn=0)
  • Uses features like contract type, monthly charges, and tenure
  • Showcases model interpretability (feature importance)

1. Run the interactive demo

  python retroml_demo.py

2. Analyze the model in detail

  python3 retroml_model_analyzer.py

3. Ensure that you have flask installed

     pip install flask

4. Launch the demo web application

  python3 retroml_web_demo.py

Then open http://localhost:5000

Project Structure

retroml/
├── retroml.py              # Main pipeline script
├── retroml_ui.py          # Interactive terminal UI (optional)
├── requirements.txt       # Python dependencies
├── configs/              # Configuration files
│   ├── classification_example.json
│   └── regression_example.json
├── data/                 # Sample datasets
│   ├── customers.csv     # Customer churn data
│   └── houses.csv        # House price data
└── results/              # Output directory
    ├── classification/
    └── regression/

Configuration Format

{
  "dataset": {
    "path": "data/your_data.csv",
    "target_column": "target"
  },
  "problem_type": "classification",  // or "regression"
  "preprocessing": {
    "impute_missing": "mean",        // "mean", "median", "mode"
    "scale_features": true
  },
  "model": {
    "auto_select": true,
    "options": ["logistic_regression", "random_forest", "xgboost"]
  },
  "output": {
    "save_dir": "results/",
    "format": "pickle"
  }
}

Features

• ✅ Automated model selection - Tests multiple algorithms automatically
• ✅ Config-driven workflow - No coding required, just JSON configuration
• ✅ Retro 90s UI - ASCII art, colored output, loading animations
• ✅ Comprehensive reporting - Detailed evaluation reports and metrics
• ✅ Model persistence - Save and load trained models
• ✅ Extensible architecture - Easy to add new algorithms and preprocessing steps

Supported Algorithms

Classification:

• Logistic Regression
• Random Forest
• XGBoost (if installed)

Regression:

• Linear Regression
• Random Forest
• XGBoost (if installed)

The Retro Experience

RetroML brings back the radical vibes of 90s computing:

• ASCII art banners and loading screens
• Colorful terminal output with Rich formatting
• Progress bars that look like old-school installers
• Nostalgic terminology and messaging

Example Usage

# Run classification pipeline
python retroml.py configs/classification_example.json

# Run regression pipeline  
python retroml.py configs/regression_example.json

# Create your own config and run
python retroml.py configs/my_config.json

Output Files

After running a pipeline, you'll find:

• best_model.pkl - The trained model (with preprocessing)
• evaluation_report.txt - Detailed performance metrics
• Console output with colorful retro styling

Extending RetroML

Want to add new algorithms or preprocessing steps? The modular design makes it easy:

1. Add new models to _get_default_models()
2. Extend preprocessing options in preprocess_data()
3. Add new problem types (clustering, NLP, etc.)

Coming Soon (Roadmap)

• Clustering algorithms (K-means, DBSCAN)
• NLP support with transformers
• Hyperparameter tuning with Optuna
• Model explainability with SHAP
• MLflow experiment tracking
• Web-based UI with Flask/Streamlit

---

"Dude, this ML pipeline is totally tubular!" - RetroML, probably

Troubleshooting

ModuleNotFoundError: Install requirements with pip install -r requirements.txt

FileNotFoundError: Make sure your dataset path in the config is correct

Performance Issues: Try reducing dataset size or using simpler models for large datasets

License

This project is open source. Feel free to modify and distribute! MIT License

---

🎮 RetroML - Bringing 90s vibes to modern ML! 🎮 Made with ❤️ by Agile Creative Labs