Data-Driven MMA Outcome Prediction Enhanced by Fighter Styles: A Machine Learning Approach

Mixed Martial Arts (MMA) analysis presents unique challenges for predicting fight outcomes and clustering fighter styles. Existing research often falls short of accurately capturing the complexity of fighters' technical styles and their impact on match results. To bridge these gaps, we propose a novel approach that utilizes machine learning methods. Specifically, 1) Use factor analysis to derive high-dimensional technical style factors and applies the K-means algorithm to cluster fighters based on these factors. 2) Various machine learning models, such as Random Forest, Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Logistic Regression, Neural Networks are experimentally tested. 3) An ensemble learning model is proposed that employs majority voting among the mentioned models. 4) Based on data from the Ultimate Fighting Championship (UFC), the experimental results demonstrate that the ensemble learning model achieves the highest accuracy of 65.52%, significantly outperforming individual models. An ablation study further validates the importance of these factors in prediction accuracy. These findings underscore the effectiveness of incorporating technical style factors into predictive models, enhancing accuracy and interpretability in the context of MMA.