|
|
--- |
|
|
license: mit |
|
|
--- |
|
|
|
|
|
# 🏀 NBA Player Performance Predictor (Assignment #2) |
|
|
|
|
|
## 🎥 Project Presentation |
|
|
[## Click here to Watch the Video ](https://youtu.be/MlvFNjFf5ZI) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
--- |
|
|
|
|
|
## 📌 Overview |
|
|
This project analyzes NBA player data (1950-2017) to predict player performance based solely on physical attributes (Height, Weight, Age). |
|
|
* **Goal 1 (Regression):** Predict exact points scored per season. |
|
|
* **Goal 2 (Classification):** Classify players as "High Scorers" vs. "Low Scorers." |
|
|
|
|
|
## 🔧 Feature Engineering |
|
|
We engineered several features to improve model performance: |
|
|
* **Clustering:** Used K-Means to identify 5 distinct body types (e.g., "Small Guard" vs. "Heavy Center"). |
|
|
* **Scaling:** Standardized Height and Weight to compare players fairly. |
|
|
* **Encoding:** Converted player positions into numeric data. |
|
|
|
|
|
## 🏆 Model Results |
|
|
|
|
|
### Part 1: Regression (Predicting Points) |
|
|
* **Winner:** Gradient Boosting Regressor |
|
|
* **R² Score:** 0.075 |
|
|
* **Insight:** Physical attributes alone are weak predictors of exact scoring numbers. |
|
|
|
|
|
### Part 2: Classification (High vs. Low Scorer) |
|
|
* **Winner:** Support Vector Machine (SVM) |
|
|
* **Recall:** 64% |
|
|
* **Insight:** The SVM model was the best at identifying "Hidden Gems" (High Recall), minimizing the chance of missing out on talent. |
|
|
|
|
|
## 📂 Files Included |
|
|
* `winning_model.pkl`: The trained Gradient Boosting Regressor. |
|
|
* `winning_classifier.pkl`: The trained SVM Classifier. |
|
|
* `My_Notebook.ipynb`: The complete Python code for this analysis. |
