The access to NBA player and team tracking data has enhanced the dimensionality of basketball. With a simple goal of putting the ball in the basket, analyzing how well a player is playing has now been transformed into something much more complicated. Luckily, Principal Component Analysis (PCA) can help reduce the dimensionality of this data to make it much more understandable for those of us who are not basketball fans. Thus, we explore the data set through the technique of PCA to understand a player’s performance and the correspondence to winning. We also transformed this to calculate a team’s PC score and evaluate a team’s performance across the NBA. After exploring the dataset, our goal was to accurately predict the relationship between a player’s statistics and the position he plays in. There are five possible positions a player can play in: point guard, shooting guard, small forward, power forward, or center. To understand which statistics correlate to position, we used feature engineering and visualizations to depict what features do help and do not help in predicting one’s position. We then created logistic regression models and random forest models to help classify the players based on these features.
Please read the project narrative for our full report.