Skip to content

cansahin9/dissertation

Repository files navigation

NBA Archetypes & Team Performance (2015–2024)

This project identifies NBA player archetypes via clustering and uses them to predict team Net Rating with machine learning. Built using 10 seasons of player/team data.

Highlights

5 Player Archetypes: Ball-Dominant Guards, 3PT Specialists, Versatile Wings, Stretch Bigs, Traditional Bigs

Models Used

GLMNet, Cubist, XGBoost, Random Forest, Stacked Ensemble

Best Model: Stacked (GLMNet + Cubist)

Explainability

SHAP highlights PER, OBPM, and Stretch Big share as key features

Repo Contents

clustering_analysis.R – Player archetype discovery

predicting_team_ntrtg.R – Team Net Rating prediction

ML_ws_bpm_cluster.R – Player BPM prediction

Other scripts for preprocessing, aggregation, and visualization

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages