UFC Fight Predictor

This project predicts the outcome probabilities of upcoming UFC fights using machine learning models trained on historical fight data.

Features

• Cleans and preprocesses UFC fight data
• Trains and evaluates Logistic Regression and Gradient Boosting models
• Calibrates model probabilities for better prediction reliability
• Predicts win probabilities for upcoming fights

How It Works

• Loads historical UFC fight data and upcoming fight cards
• Cleans and processes the data, engineering features for model input
• Splits data into training and validation sets, removes outliers, caps and scales features
• Trains and tunes models using GridSearchCV
• Calibrates the best model and predicts win probabilities for new fights

Usage

1. Clone this repository:

   git clone https://github.com/yourusername/ufc-fight-predictor.git
   cd ufc-fight-predictor

2. Install dependencies:

   pip install -r requirements.txt

3. Place your data files (ufc-master.csv and ufc_fight_cards.csv) in the project directory.
4. Open and run the Jupyter notebook alternate_predictor.ipynb:

   jupyter notebook alternate_predictor.ipynb

Requirements

• Python 3.8+
• numpy
• pandas
• scikit-learn
• jupyter

Install all dependencies with:

pip install -r requirements.txt

Notes

• Data files are not included due to size. Please provide your own ufc-master.csv and ufc_fight_cards.csv.
• The notebook is well-commented for easy understanding and modification.

License

MIT License