Predictions of Basketball Match Results using Autoencoder with Gated Recurrent Unit Algorithm

This research explores basketball match predictions within the field of sports analytics utilized Deep Learning (DL) technique to analyse the outcomes of National Basketball Association (NBA) games based on player performance and team statistics. The challenge in predicting NBA games outcomes and performance analysis various factor, as the high dimensionality of the data and uncertainty surrounding player due to difficult to predictions of basketball game. This paper, proposed Autoencoder with Gated Recurrent Unit (AE-GRU) algorithm to efficiently analyse and predict NBA game outcomes. The data was initially collected from NBA games between 2009 - 2021, focusing on five teams and pre-processing involved min-max normalization to uniformly scale the feature vectors. The Principal Component Analysis (PCA) utilized for feature selection, effectively reducing the high dimensionality of the NBA game data across various teams. The AE-GRU technique then captured the complex feature of basketball game data and handled uncertainty data and enhances performance. The performance of DL model evaluated the various metrics such as accuracy, precision, recall and f1-score metrics on NBA game.