Machine learning and feature representation for predicting NCAA division I basketball tournament outcomes
Publisher
Andres Gregori Altamirano
Date Issued
2026
Author(s)
Gregori Altamirano, Andres
Advisor(s)
Escobar Castillejos, David
Barrera Ánimas, Ari Yair
Type
text::thesis::master thesis
Abstract
Este estudio evalúa el impacto del Análisis de Componentes Principales (PCA) en la predicción del torneo de baloncesto de la NCAA usando los modelos SVC, Random Forest y XGBoost. Con datos de 2003 a 2025, PCA-SVC logró el mejor desempeño (F1 de 84.33%), concluyendo que su beneficio es específico de cada clasificador
License
Acceso Abierto
How to cite
Gregori Altamirano, A. (2026). Machine learning and feature representation for predicting NCAA division I basketball tournament outcomes. (Tesis de maestría). Universidad Panamericana.
Table of contents
Chapter 1. Introduction -- Background -- Motivation -- Problem Statement -- General Objective -- Particular Objectives -- Hypothesis -- Thesis Structure -- Chapter 2. Related Work -- Machine Learning in Sports Prediction -- Prediction Tasks and Data Sources -- Models Commonly Used in Sports Analytics -- Basketball Outcome Prediction -- March Madness and Tournament Forecasting -- Feature Representation and Dimensionality Reduction -- Validation and Evaluation Practices -- Position of the Present Study -- Chapter 3. Methodology -- Research Design -- Dataset -- Outcome Definition -- Feature Construction -- Data Preprocessing -- Principal Component Analysis -- Machine Learning Models -- Training and Evaluation Protocol -- Chapter 4. Results -- Overview of the Evaluation -- Model Performance -- Effect of PCA by Model -- Comparison Across Feature Representations -- Summary of Findings -- Chapter 5. Discussion -- Interpretation of Results -- Relation to the Literature -- Interpretability -- Limitations -- Future Work -- Chapter 6. Conclusion -- Appendix A. Supplementary Feature Dictionary