The Use of Machine Learning in Predicting Formula 1 Race Outcomes

Formula 1 is a sport driven by both engineering excellence and data precision. With an ever-growing wealth of historical and real-time race data, machine learning offers an opportunity to transform race outcome prediction and strategy development. Prior literature in motorsport analytics highlights the use of classification and regression models, yet few studies leverage deep learning models specifically built for structured data. This study aimed to develop a robust machine learning pipeline to predict both driver finishing positions and constructor championship points using historical F1 data from 2010 to 2023. TabNet, a deep learning architecture optimized for tabular data, was selected for its interpretability and strong feature selection capability. The models were trained using pre-race variables such as grid position, number of laps, constructor, and overtakes. Hyperparameter tuning was conducted using Optuna. The results showed strong predictive performance, with the driver model outperforming the constructor model in overall accuracy. These findings demonstrate the practical potential of machine learning in high-stakes motorsport environments. The models developed in this research could support teams in strategic planning, broadcasters in providing predictive insights, and analysts in simulating race outcomes under different conditions.