NBA Salary Prediction Using Machine Learning

COSC 325 — Spring 2026 Course Project

Overview

This project uses supervised machine learning to predict 2025–26 NBA player salaries (in millions of dollars) from performance statistics. We scrape advanced stats, per-game stats, and contract data from Basketball Reference, then train and compare multiple regression models to see which stats the NBA actually pays for.

Dataset

• Source: Basketball Reference (scraped via playerDataSet.py)
• Players: 438 (after filtering to ≥10 games played)
• Features: 28 (24 numeric stats + 4 position dummies)
• Target: Salary in $M (log-transformed for modeling)

Models

Model	Status	R²
Median Baseline	Done	-0.208
Linear Regression	Done	0.507
Ridge Regression	Done	0.520
Decision Tree	Done	0.622
Random Forest	Done	0.678
MLP (Neural Network)	Done	0.500

Repo Structure

• playerDataSet.py — Scrapes and cleans data, generates EDA plots
• baseline_salary_prediction.ipynb — Preprocessing, models, and evaluation
• nba_salary_2025_26.csv — Cleaned dataset (438 players, 29 columns)

Still To Do

• Decision Tree with overfitting analysis
• Random Forest with feature importance analysis
• MLP (neural network) implementation and tuning
• Final model comparison table and plots
• Presentation slides (due Apr 29)
• Final report — 4-page IEEE format (due May 6)

How to Run

1. Install dependencies: pip install pandas numpy scikit-learn matplotlib seaborn
2. Open baseline_salary_prediction.ipynb in VS Code or Jupyter
3. Place nba_salary_2025_26.csv in the same directory
4. Run All