Multi-Dimensional Clustering of Roles in the NBA
While in the National Basketball Association (NBA), players are often described by the position that they play and not necessarily the role that they fill on the team. In this thesis, newly defined player roles have been identified by applying multi-dimensional clustering techniques on thirty-eight...
| Autor: | |
|---|---|
| Tipo de documento: | dissertação |
| Estado: | Versión aceptada para publicación |
| Data de publicação: | 2021 |
| País: | México |
| Recursos: | Instituto Tecnológico y de Estudios Superiores de Occidente |
| Repositório: | Repositorio Institucional del ITESO |
| Idioma: | inglês |
| OAI Identifier: | oai:rei.iteso.mx:11117/7441 |
| Acesso em linha: | https://hdl.handle.net/11117/7441 |
| Access Level: | Acceso aberto |
| Palavra-chave: | Clustering Multi-Dimensional Clustering Gaussian Mixtures DBSCAN k-Means |
| Resumo: | While in the National Basketball Association (NBA), players are often described by the position that they play and not necessarily the role that they fill on the team. In this thesis, newly defined player roles have been identified by applying multi-dimensional clustering techniques on thirty-eight variables for over ten thousand player samples. These roles help to differentiate players that play the same traditional position, and will allow for new comparisons between players to be produced. Using player statistics from nineteen seasons, models were developed using three separate clustering techniques: Gaussian Mixtures, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and k-Means. After the models were developed a final model was chosen that provided the best clusters that were used to identify the new roles. These new roles are able to be used to identify replacements for certain players, signing a player that fulfills the same role, or by drawing comparisons between new players in the NBA and the historical roles that other players have fulfilled. |
|---|