Badminton match outcome prediction model using naive bayes and feature weighting technique.

Abstract

The recent growth in the field of data mining and machine learning has remitted into more recognition of outcome prediction and classification. However, the application of these techniques in the field of sports is still unexplored. This paper presents the implementation of data mining and machine learning in sports particularly. Here, machine learning based algorithm to predict the outcome of the badminton tournament has been proposed. We have employed three classifiers, Naïve Bayes with Correlation Based Feature Weighting (NB-CBFW), Composite Hypercubes on Iterated Random Projections (CHIRP) and Hyper Pipes to predict the outcome of Australian Open 2019, Malaysian Open 2019, German Open 2019 and Singapore Open 2019 Badminton tournaments. The outcome prediction is measured in terms of Accuracy, Root Mean Square Error (RMSE), True Positive Rate (TPR), True Negative Rate (TNR), Positive Predicted Value (PPV), Negative Predicted Value (NPV) and Receiver Operating Characteristics (ROC). After implementing the classifiers, it has been observed that NB-CBFW shows excellent accuracy in match outcome prediction as compared to CHIRP and Hyper Pipes.