🚖 Taxi Trip Total Fare Prediction (Multi-Model Ensemble)
An end-to-end Machine Learning regression project built to predict the total fare amount for taxi rides. This project evaluates and compares four powerful algorithms to find the most accurate and reliable architecture for fare forecasting.
📊 Model Performance & Evaluation ($R^2$ Scores)
The models were evaluated based on their coefficient of determination ($R^2$), delivering highly impressive results:
- Linear Regression (LR): 99.64% accuracy (Best overall structural fit)
- Gradient Boosting Regressor (GB): 99.44% accuracy (Exceptional ensemble learning)
- XGBoost Regressor (XG): 98.89% accuracy (Highly optimized gradient boosting)
- K-Neighbors Regressor (KNN): 97.23% accuracy (Robust distance-based baseline)
🚀 Key Features
- Multi-Algorithm Benchmark: Provides a transparent performance comparison across linear, distance-based, and tree-based ensemble models.
- Exceptional Precision: Achieves up to 99.64% accuracy by properly learning from features like trip distance, location patterns, and timestamps.
- Production Ready: The best-performing pipelines are serialized and structured for instant inference integration into ride-hailing software.
📁 Repository Structure
models/: Directory containing serialized model files (LR, GB, XG, KNN saved via pickle/joblib).preprocessor.pickle: Serialized scaling and encoding pipeline for structural alignment.train.py / notebook.ipynb: Core Python code handling data cleaning, feature engineering, training, and evaluation.app.py: UI interface file for Hugging Face Spaces.README.md: Comprehensive documentation.
💻 Tech Stack
- Language: Python
- ML Libraries: Scikit-Learn, XGBoost
- Data Handling: Pandas, NumPy
- Deployment: Hugging Face Spaces & Gradio
📜 License
This project is licensed under the MIT License.
