Predictive Analytics in NBA Performance Using Piecewise and Spline Regression Models
Python | Data Manipulation | Data Visualization (pandas, numpy, matplotlib, seaborn) | Statistical Modeling (Linear, Polynomial, Piecewise, Spline Regression) | Data Imputation (Simple, Iterative, KNN, LGBM Imputation) | Experiment Tracking (ML Foundry) | Model Evaluation (RMSE, MAE, MSE, R², ANOVA) | Machine Learning (Scikit-learn for regression, model fitting, and evaluation) | Data Visualization (seaborn, matplotlib) | Model Fitting | Statistical Significance Testing
Project Description
This project focuses on developing advanced regression models to predict the points scored by NBA teams based on various fitness and performance metrics such as weightlifting sessions, yoga sessions, water intake, and laps run during practice. Utilizing techniques like Piecewise and Spline Regression, I aimed to identify the most effective model to represent the non-linear relationships between performance factors and scoring output.
Project Skills
Linear Regression: Built baseline models to predict points scored using traditional linear regression techniques.
​
Polynomial Regression: Employed quadratic and cubic models to account for non-linear relationships between variables.
​
Piecewise Regression: Applied a piecewise function to segment data into intervals and modeled each section independently for better predictive accuracy.
​
Spline Regression: Introduced spline regression to smoothen transitions between segmented models, providing a more accurate representation of complex relationships.
​
Hypothesis Testing:
Conducted T-tests to determine the statistical significance of individual predictors.
Used ANOVA (Analysis of Variance) to compare between-group variances and validate the model’s performance across different teams.
​​
Data Cleaning & Processing:
Implemented outlier detection and removal using interquartile ranges.
Applied missing data imputation using KNN Imputation, Iterative Imputer, and other advanced techniques, ensuring the integrity of the dataset.
​
Data Visualization:
Utilized Seaborn and Matplotlib to generate visual insights, including density plots, box plots, and histograms, revealing data distribution, outliers, and trends.
Model Evaluation:
Evaluated model performance using metrics like Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared. Performed a comprehensive comparison of different models to select the best performing one for the business task.
​
​
Project Demostrates
This project demonstrates a high level of proficiency in data analytics, economic modeling, and strategic insights by integrating data from public sources and employing sophisticated methods to analyze wealth trends. His understanding of the intersection between economic policies and wealth distribution reflects his ability to use data-driven approaches to inform business strategy and economic decision-making. The project highlights Joseph's unique skill set in data manipulation, geospatial analysis, and economic forecasting, positioning him as a valuable asset in addressing complex challenges in the intersection of economics and data analytics.