Deep Learning to Reliable Score Prediction in Hundred Ball Cricket Matches

Cricket is an unpredictable sport in which two teams compete. The most challenging task in cricket is to predict possible cricket scores, where reliability (in terms of accuracy of score prediction) is crucial. Traditional approaches and methods are limited regarding reliability and timely score prediction, particularly in the case of hundred-ball cricket to forecast probable scores. In this paper, we study cricket game events and related datasets and perform the score prediction of cricket matches. To make the match prediction, we employ a deep learning method, which is not yet applied in cricket, especially in hundred-ball cricket, to forecast probable scores. In the deep learning method, we apply a Multi-layer Perceptron (MLP) network and design a Multi-layer Perceptron Regression to develop several models for the score prediction. We perform predictive analysis on a live hundred-ball cricket competition to forecast the ultimate score of the game before the match begins and/or on running matches. The evaluation shows that the proposed MLP Regression model can provide up to 99% accuracy in score prediction, indicating that the proposed deep learning-based method is reliable and flawless.