NBA Player Clustering
“2013 NBA Logo” by RMTip21 is licensed under CC BY-SA 2.0.
A machine learning framework to reclassify NBA players beyond traditional positions using 10 years of player data (2014-2024).
I applied K-means clustering with PCA dimensionality reduction to identify 5 distinct player archetypes: Stars, Snipers, Painters, Shooters, and Protectors.
Key Technical Components:
- Processed 10 seasons of NBA advanced statistics, per-100 possession metrics, and shooting data
- Implemented weighted averaging based on minutes played to capture career performance
- Evaluated 20 clustering models using multiple validation metrics (Silhouette Score, Calinski-Harabasz Index)
- Used t-SNE visualization for cluster interpretation and radar charts for feature analysis
Business Application:
- Applied clustering framework to Chicago Bulls roster analysis
- Identified team composition gaps and recommended specific free agent acquisitions
- Built OLS regression model predicting team wins with 91.8% R-squared accuracy
- Analyzed four-factor model (shooting, turnovers, rebounding, free throws) to evaluate team performance