NBA Vol. II

The purpose of this project is to demonstrate the application of various predictive models on a given dataset. This project have a bunch of csv files already preprocessed in NBA Vol. I to use as input datasets. The project uses Jupyter Notebook for exploratory data analysis, feature engineering, and model development.

This project utilizes scikit-learn 1.2.2 a widely used and popular Python library, as the core framework for implementing various machine learning models and algorithms.

Goals

The goals of this project are:

1. Define a prediction task
2. Select evaluation metrics and baseline models
3. Perform feature engineering and standardization
4. Train and use predictive models: Univariable and Multivariate Linear Regression, Classification
5. Understand how Gradient Descent works by implementing a Linear Regressor in Python

Data

NBA Season 21/22. You will find everything you need here NBA Vol. I.

Methods

The following methods were used in this project:

1. Exploratory Data Analysis
2. Feature Engineering
3. Standardization
4. Univariable Linear Regression
5. Multivariate Linear Regression
6. Classification
7. Gradient Descent

Results

1. Salary prediction (continuous variable): the MAE dropped almost by 50% between Baseline and Decision Tree.

   • MAE Baseline: $ 6.948.493
   • MAE Decision Tree: $ 3.780.848

2. All Star selection (discrete variable / Logistic regression):

   • Baseline metrics: a) F1: 0.724 b) Recall: 0.669 c) Precision: 0.788 d) Accuracy: 0.984
   • After fine-tuning and feature engineering: a) F1: 0.769 (↑ 6%) b) Recall: 1 (↑ 49.47%) c) Precision: 0.625 (↓ 26%)

How to Use

1. Remember, before installing all the dependencies with !pip install -r requirements.txt, build a virtual/working env to avoid conflict between your local installations and the packages used in this project.

Conclusion

To summarize, this project showcases the successful implementation of several predictive models on a specific dataset, achieving the predefined objectives and generating valuable insights. The project's outcomes can serve as a useful reference for similar predictive modeling tasks and highlight the practical application of machine learning in data analysis and decision-making