Score and winning prediction in cricket through data mining

Currently, in One Day International (ODI) cricket matches first innings score is predicted on the basis of Current Run Rate which can be calculated as the amount of runs scored per the number of overs bowled. It does not include factors like number of wickets fallen and venue of the match. Furthermore, in second innings there is no method to predict the outcome of the match. In this paper a model has been proposed that has two methods, first predicts the score of first innings not only on the basis of current run rate but also considers number of wickets fallen, venue of the match and batting team. The second method predicts the outcome of the match in the second innings considering the same attributes as of the former method along with the target given to the batting team. These two methods have been implemented using Linear Regression Classifier and Naïve Bayes Classifier for first innings and second innings respectively. In both methods, 5 over intervals have been made from 50 overs of the match and at each interval above mentioned attributes have been recorded of all non-curtailed matches played between 2002 and 2014 of every team independently. It has been found in the results that error in Linear Regression classifier is less than Current Run Rate method in estimating the final score and also accuracy of Naïve Bayes in predicting match outcome has been 68% initially from 0-5 overs to 91% till the end of 45th over.