Estimation of American Basketball League (NBA) Match Results by Artificial Neural Networks

In this study, American National Basketball League (NBA) match results were tried to be predicted using artificial neural networks by using 2015-2016 season data. This prediction can be defined as a classification because it will take 0 to 1 value as the win or lose. In the study, 596 matches selected from different matchday of the NBA 2015-2016 season were modeled to have 11 different input parameters and 1 output parameter. 396 of them were trained and 200 of them were randomly selected as test data, and an artificial neural network with 3 hidden neurons was designed on both two layers. Logarithmic sigmoid was used as transfer function and purelin function was used as output function, and the values produced by the network are rounded to zero or one for clustering as win or lose. Artificial neural networks have performed well in this model, which is about 90% success. Moreover, by examining the confusion matrix, it has been shown that network faults are distributed regular