Gabriel Reynoso Romero

Projects

Exploring prediction, optimization, and real-world problems through data and engineering.

View GitHub Profile

FIFA World Cup Forecasting

A two-stage probabilistic forecasting system for the 2026 World Cup, combining a Dixon-Coles match model with a LightGBM calibrator and a Monte Carlo tournament simulation.

  • Python
  • Dixon-Coles model
  • LightGBM
  • scikit-learn
  • Elo ratings
  • Monte Carlo simulation
  • 0.8711 test log-loss, +20.7% vs. uniform baseline
  • Evaluated on 4,552 held-out matches (2022+)
  • 100,000-run Monte Carlo tournament simulation
Code

Traffic Optimization via Collective Intelligence

Comparing greedy routing, marginal-cost assignment, and a Q-learning agent trained by expert imitation on a real Manhattan street network.

  • Python
  • Reinforcement learning (Q-learning)
  • Graph algorithms
  • NetworkX
  • Jupyter
Code

Book Recommendation System

A desktop book-recommendation app that clusters books with k-means and suggests similar titles by cosine similarity, built with a three-person team as a course project.

  • Python
  • Tkinter
  • scikit-learn
  • k-means clustering
  • Cosine similarity
  • pandas
Code

Statistical Analysis of La Liga Shot Events

A statistical analysis of shot event data from La Liga's 2017/18 season, using hypothesis testing and logistic regression to examine which factors are associated with a shot becoming a goal.

  • Python
  • pandas
  • SciPy
  • statsmodels
  • scikit-learn
  • Logistic Regression
Code