NBA Playoff Qualification: A Hybrid Machine Learning and Social Network Analysis Approach

In recent years, sports data analysis has emerged as a rapidly growing research field, becoming one of the core aspects of modern sports management. It can assist coaches and sports managers in predicting match outcomes and evaluating team and player performance. One of the most popular sports competitions, the NBA basketball games, has been the subject of numerous studies where various machine learning (ML) models have been used to predict game outcomes. However, previous research has primarily focused on traditional machine learning models, often overlooking the complex relationships between players within a team and their performance in interaction with other teammates. Therefore, this study proposes a hybrid model to predict NBA teams' qualification for the playoffs, utilizing complete game data from the 2007-2008 to 2017-2018 seasons. The experiments in this study were based on three different data scenarios: team statistical features on a per-game basis, purely network-based features, and finally, a combination of statistical and network features of the games. The results of this study indicate that the use of the hybrid approach significantly increased prediction accuracy across various models. Furthermore, in the hybrid approach, both the Logistic Regression and XGBoost models, with an accuracy of 93%, demonstrated the best performance compared to other models. The findings of this study provide valuable insights for future research in predicting and analyzing the performance of basketball teams.