angel-721/ufp

The Ultimate Fight Predictor

UFP

The Ultimate Fight Predictor- A fun way to predict the outcome of UFC fights.

Warning(For now)

The current models were trained on scikit-learn 1.20 so predicting on later versions might not work if you are on a different version of scikit-learn

• model1.pkl is trained on the latest version of scikit-learn!

Use

Install requirements with pip install -r requirements.txt Enter fighter data in the way specified by ./data/example.csv and run main.py with the data you have as a command line argument and predict as the action.

For example just run python main.py and it should use model1.pkl to predict everything in example.csv.

Data for training

The model was trained on the Ultimate UFC Dataset. If you want to train your own model with the ground work please download the data first.

Background information

• model1.pkl is trained on a SVM using feature selection
• The problem is a binary classification where the label is Winner which can be Blue or Red
  • The features are described in the dataset and include some aspects such as betting odds, wins, losses, and height
• All of the training is done on a 80/20(train/test) split with cross validation on 15 folds