Leveraging Graph Neural Networks and Gate Recurrent Units for Accurate and Transparent Prediction of Baseball Pitching Speed

Abstract

Long Short-Term Memory (LSTM) networks are widely used in biomechanical data analysis but have significant limitations in interpretability and decision transparency. Combining Graph Neural Networks (GNNs) with Gate Recurrent Units (GRUs) may offer a better solution. This study proposes and validates a hybrid GNN-GRU model for predicting baseball pitching speed, enhancing its interpretability using Layer-wise Relevance Propagation (LRP). C3D data from 53 baseball players were downloaded from a public dataset. Kinematic features of 9 joints and pitching speed during the pitching process were calculated using Visual3D, resulting in a total of 208 valid pitches. The feature data were input into both LSTM and GNN-GRU hybrid models, with hyperparameters tuned using Particle Swarm Optimization (PSO). LRP was employed to obtain the contribution rate changes of kinematic features to the prediction results throughout the pitching cycle. The prediction accuracy of the models was evaluated using Mean Absolute Error (MAE) and R-squared (R²). The results showed that there were significant statistical differences in the MAE and R2 metrics between the LSTM model and the GNN-GRU model in predicting pitching speed on the test set. The MAE (P = 0.000, Z = -5.170, Cohen's d =1.514) and R2(P = 0.000, Z = -2.981, Cohen's d =2.314) of the LSTM model were significantly lower than those of the GNN-GRU model. Compared to LSTM, the GNN-GRU model achieved better prediction accuracy but was more susceptible to the influence of data outliers. Moreover, the GNN-GRU-based model demonstrated better interpretability and decision transparency.