Predicting Basketball Results Using Cascading Algorithm

Anybody can guess the winners of a basket game. The question is how big the chances are in predicting the real winners. Relying only on the experts' experiences and intuition could not discover all the value and potential of the collected data. Driven by the increasing comprehensive data in sports datasets and data mining technique successfully used in different areas, sports data mining technique emerges and enables us to find hidden knowledge to impact the sports industry. A more scientific approach is needed to use for these data that are collected. Some predictors based only on winning records and some based only on statistical records of both teams. There are also predictors which use both types of data, but the accuracy of applying different individual algorithms is only ranging about 60% - 70%. To achieve better prediction rates and deal with that complexity, a lot of machine learning methods have been implemented over these data. This paper presents an improved technique for predicting basketball game results implementing cascading algorithm. The researchers combined Naive Bayes, Four Factor Analysis, and Fuzzy Logic Algorithms to predict basketball game result in an acceptable level of 69% - 70% accuracy. The researchers tested several times using data sets from NBA game Season 2015-2016, and the cascading algorithm result manages to reach 70% prediction accuracy. The result of this system can be used to assist basketball coaches in making plans for possible team developments. Also, the forecasted results can serve as an aid in building effective gameplay.