The Application of Machine Learning in League Index Prediction

Abstract

Along with the technological developments and the commercial utilization of 4G, electronic sports and live broadcasting institutions have got huge progresses. Thus, it is vital and meaningful to optimize the gaming circumstances and maximize the gaming experiences of players and this is why exploring the main contributing factors or players' league indexes as well as constructing predictive models are so necessary and pragmatic. It is the experiences and the observations that the traditional methods applied for judgments, which cannot handle massive data and will end up with low-accuracy outcomes, whereas the situations can be processed effectively, precisely and objectively via machine learning. In this paper, via processing over 3338 data of 19 variables extracted from real game players, descriptive statistical analysis has been firstly processed for identifying the real factors that influence the players' league indexes, then, six well-known machine learning models are used to build the prediction models. We have discovered that, during the tests, Artificial Neural Network model offers the best prediction and correctly predicts over 45.4% of the testing data.