Relative Analysis and Performance of Machine Learning Approaches in Sports

Machine Learning involves conceiving algorithms that enable computing systems to recognize patterns in information set by employing statistical tools to make predictions. In Machine Learning a grouping model is generated from the data set for forecasting test data nature through training data. From Financial institutions, E-commerce marketplace and stock markets to health care domain, day to day human activity automation and wildlife conservation, every field is now exploring machine learning for achieving robust, accurate and optimized solutions. Linear Regression, Logistic Regression, Decision Trees, Random Forest, K-Means Clustering, Artificial Neural Network, Deep Learning Architectures are some of the algorithms in Machine Learning, with distinct characteristics for being used as a classification model or a regression model. In this work, three popular machine learning algorithms namely random forest, Support vector machines (SVM) and multi-layer perceptron's are employed to analyze and predict the performance in the game of cricket. Dataset was prepared by collecting live scores from IPL T20 happened in 2019, 2020 and so on. The accuracy is found to increase with the increased input data set for training and optimization based on learning rates and vector parameters. The model is characterized based on the player abilities and their performances which play an eminent role in final prediction. From the study it is observed that Multi-layer perceptron's are able to outperform Random forest and SVM.