Predicting the outcome of badminton matches based on network analysis

The expansion of World Wide Web and faster access to the Internet is changing its primary use. Websites contain a lot more information that can be acquired and processed. Such sites are also the ones that provide results for various competitions in sport. Dealing with sport statistics is well spread among fans and journalists, as well as sport analysts. The purpose of this thesis is the prediction of rank and total points scored by badminton players in Slovenia on a virtual ranking that is based on tournament results from a certain time window. Prediction is based on attributes that are obtained with the analysis of a network representing all games played in this period. The analysis is based on the data available through www.tournamentsoftware.com. The acquired data has first been processed and represented in the form of a network using Java programming language and JUNG network analysis library. The necessary attributes were obtained using the techniques from network analysis and then adopted to learn a prediction model. The mentioned attributes were the basis for running various statistical, data mining and machine learning methods in Orange. The main contribution of the thesis is a model that can be used for predicting results and standings in badminton in general or in any other sport that runs competitions or tournaments in a similar manner as badminton (with a certain adjustments). Due to a small sample size (the number of regular players in Slovenia), the performance of the model in practical applications remains unclear