A Comparative Study of Machine Learning Algorithms for Prior Prediction of UFC Fights

Mixed Martial Arts is a rapidly growing combat sport that has a highly multi-dimensional nature. Due to a large number of possible strategies available to each fighter, and multitude of skills and techniques involved, the potential for upset in any fight is very high. That is the chance of a highly skilled, veteran athlete being defeated by an athlete with significantly less experience is possible. This problem is further exacerbated by the lack of a well-defined, time series database of fighter profiles prior to every fight. In this paper, we attempt to develop an efficient model based on the machine learning algorithms for the prior prediction of UFC fights. The efficacy of various machine learning models based on Perceptron, Random Forests, Decision Trees classifier, Stochastic Gradient Descent (SGD) classifier, Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) classifiers is tested on a time series set of a fighter's data before each fight.