Treballs Fi de GrauEnginyeria Informàtica i Matemàtiques

Analysis and prediction of football results with artificial intelligence: a practical application of temporal fusion transformers

  • Identification data

    Identifier:  TFG:9470
    Authors:  Garcia Blas, Joan
    Abstract:
    The project applies advanced AI to predict football results in the betting field. It implements a Temporal Fusion Transformer capable of generating multivariate predictions (goals, corners, cards) with high precision and interpretability, integrated into a modern architecture (FastAPI, PostgreSQL, React). The platform allows for specific and day-by-day predictions and offers a searchable history. Backtesting over the last 10 days (100 matches) achieves a 50% 1X2 accuracy; it underestimates goals, shots and corners and overestimates yellow cards. Low response times and consistency validate the system. Future: better calibration, more context, real time and expansion.
  • Others:

    Access rights: info:eu-repo/semantics/openAccess
    Education area(s): Enginyeria Informàtica
    Department: Enginyeria Informàtica i Matemàtiques
    Entity: Universitat Rovira i Virgili (URV)
    Confidenciality: No
    Subject: Intel·ligència artificial
    Project director: Massaguer Pla, Jordi
    Work's public defense date: 2025-09-08
    Creation date in repository: 2026-06-26
    Academic year: 2024-2025
    Student: Garcia Blas, Joan
  • Keywords:

    Multivariate time series
    Football match outcome prediction
    Computer engineering
  • Documents:

  • Cerca a google

    Search to google scholar