Advanced Predictive Modeling of Sports Performance Using Google Prediction API and Ensemble Learning Framework

This study addresses the need of to provide predictive modeling systems for sports performance addressing limitations in existing methods that often lack precision and scalability. Using a Long Short-Term Memory network and Gradient Boosting Machines will predict the team's and athlete's performance metrics. The dataset provides plenty of information to train and evaluate models, with over 200,000 records in the years 2015–2023. It includes athlete-specific metrics, team-specific metrics, and factors that are outside their control. The results from the proposed procedure were outstanding as the key performance metrics all exceeded 97%. The prediction powers were shown to be better when compared to other models, like RF and LSTMs. Statistical tests, including an AUC of 0.98 and a mean absolute error of 1.2%, gave further proof of the performance of the model.