Selection of Players and Team for an Indian Premier League Cricket Match Using Ensembles of Classifiers

In In this work, we have shown how neural network and K-Means or Hierarchical clustering can be used in a combination, where clustering algorithms were used to significantly represent data in a meaningful way to predict the best batsman/bowler who is to be sent next for a given match condition at an Indian Premier League Cricket Match. We predicted the best bowler/batsman for a given match condition using match parameter of each and every ball. Tests were carried out using a variety of neural networks ranging from single layered perceptron to multilayered multi-neuron neural networks along with variety of data representation. Among these, the neural network with 3 hidden layers perceptron gave best results for batsman and bowlers respectively. These neural networks can produce fast and accurate results within 3 seconds. The activation functions used were ReLu and SoftMax. K-Means and hierarchical clustering was used to generate data. We also predict an ideal team for a match using K-Means Clustering and hierarchical clustering which found interesting and accurate patterns, constraints.