Piecewise and Spline Regression using NumPy - Predictive Sports Analytics

Overview

This project focuses on forecasting the points earned by a sports team based on their training performance, including factors like yoga sessions, laps completed, water intake, and weightlifting sessions. By applying machine learning techniques, this project aims to assist stakeholders in optimizing the team's training performance.

In this project, we explore Piecewise Regression and Spline Regression techniques to improve upon our previous results. Piecewise Regression divides the independent variable into intervals and fits different linear functions to each interval. Spline Regression combines polynomial regression functions into a piecewise continuous function.

---

Aim

The main goal of this project is to predict a sports team's points using Piecewise and Spline Regression.

---

Data Description

The dataset used in this project contains information about the points scored by sports teams based on various attributes.

• NBA_Dataset_csv.csv

---

Tech Stack

• Language: Python
• Libraries: pandas, numpy, scipy, matplotlib, seaborn, sklearn, statsmodels, piecewise_regression, csaps, py-earth, mlfoundry

---

Approach

The project follows a structured approach, including the following steps:

1. Data Reading
2. Data Preprocessing
   • Outlier Removal
   • One-Hot Encoding
   • Imputing Missing Values
3. Model Building
   • Linear Regression
   • Polynomial Regression
   • Step Functions
   • Piecewise Regression
   • Basis Functions
   • Spline Regression
     • Univariate Model
     • Bivariate Model
     • Multivariate Adaptive Regression Splines (MARS)
4. Model Evaluation and Comparison
5. Experiment Tracking with ML Foundry

---

Modular Code Overview

1. input: Contains the NBA dataset used in the project.
2. lib: A reference folder containing the original Jupyter notebook.
3. ml_pipeline: Contains functions divided into different Python files, which are appropriately named. The Engine.py script calls these functions to execute the project's steps, train the model, and display the results.
4. requirements.txt: Lists all the required libraries along with their versions. You can install these libraries using the command pip install -r requirements.txt.

---

Execution Instructions:

• Install requirements with "pip install -r requirements.txt"
• Run Engine.py to train models and get the results

---