Player's performance prediction in ODI cricket using machine learning algorithms

Abstract

This paper presents a method that is aimed towards predicting a cricket player's upcoming match performance by implementing machine learning algorithms. The proposed model consists of statistical data of players of Bangladesh national cricket team which has been collected from trusted sports websites, feature selection algorithms such as recursive feature elimination and univariate selection and machine learning algorithms such as linear regression, support vector machine with linear and polynomial kernel. To implement the proposed model, the accumulated statistical data is processed into numerical value in order to implement those in the algorithms. Furthermore, aforementioned feature selection algorithms are applied for extracting the attributes that are more related to the output feature. Additionally, the machine learning algorithms are used to predict runs scored by a batsman and runs considered by a bowler in the upcoming match. The experimental setup demonstrates that the model gives up to 91.5% accuracy for batsman Tamim and up to 75.3% accuracy for bowler Mahmudullah whereas prediction accuracy for other players are also up to the mark. Therefore, this will help in calculating player's future performance and thus will ensure better team selection for forthcoming cricket matches.