Abstract
The goal of this study was to create a decision-support system for establishing the best team roster for basketball using machine learning and MCDM techniques. The research was performed using the statistical analysis of data pertaining to the fifty best collegians who played in the 2024-2025 NBA regular season, as reported by ESPN. The players were assessed on ten separate performance metrics such as: points per game, field goal percentage, three-point shooting percentage, free-throw shooting percentage, total rebounds – total and defensive rebounds, assists, turnovers, steals and blocks. The first phase used the Two-Step clustering algorithm to categorize the players into five groups based on performance criteria, or traditional basketball positions. The second phase employed RAWEC for player rankings in each cluster based on performance metrics to identify the five highest ranked players to start for that team. Ultimately an analytical, data-driven decision-support system model was created, providing an objective means to identify players for starting positions, which will assist in the sports industry with player analysis, roster design and prediction of players’ future performance. Overall, this study provides evidence that the use of AI based models will significantly increase the efficiency of decision making in professional sports.
Keywords: Basketball Team Selection, Multi-Criteria Decision-Making (MCDM), RAWEC, Two-step Clustering, Machine Learning.