Deep Learning Approach for Forecasting Athletes' Performance in Sports Tournaments

Sports and international tournaments have gained world attention in the past decade. Enhancing sports activities and promoting sports to participate in international events, competitions, and tournaments play a substantial role in the development and advancement of nations around the globe. In this paper, we applied different deep learning models for predicting athletes' performance in tournaments to help them improve their results. We propose a deep learning selection algorithm to evaluate the effectiveness of the athletes' current training by predicting their race results upon completing each additional training, which potentially improves their performance. We gathered public training data for athletes who participated in the 2017 Boston Marathon within a five-month window prior to the race. Deep learning models were applied and evaluated to predict marathon finishing times. These include Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU). Results show that Deep Learning models give improved race time prediction accuracy over the baseline machine learning model, such as standard Linear Regression (LR).