Applying Deep Learning and Machine Learning to UFC Fights
Abstract - MMA is one of the most popular sports in the world. UFC's popularity has skyrocketed in the past few years. Moreover, as a sport, people naturally started to bet money on which fighter would win. Betting rates were set; whoever was more advantageous was called the topdog, and whoever was more likely to lose was called the underdog. As it is now the fourth industrial revolution, it is possible to apply machine learning and deep learning to predict the result of fights. When using deep learning to predict the winner, many models, including LGBM, Random Forest, XGB, Gradient Boosting, Logistic Regression, Decision Tree, Extra Tree, or SVM, were used to predict, and Logistic Regression scored the highest accuracy rate of 65.35%. Even though 65.35% is not a very high rate, 65.35% is still a very high accuracy rate compared to other studies. The data used in the study includes information on betting rates, fighter’s name, their value, fighting place, or nationality, which can help deep learning. Therefore, this study includes applying deep learning to develop a recommendation system based on a fighter. The SVD method allows the program to recommend other fighters similar to a given fighter. For instance, Alexander Volkanovski(Featherweight Champion) or Dan Ige(Featherweight Ranking #10) was recommended as a similar fighter based on Korean Zombie Chan Sung Jung(Featherweight Ranking #4). In later studies, our goal is to higher the accuracy rate by considering the matchups between fighters.