Research on Running Data Analysis Method Based on Attention-LSTM

Abstract

Predicting injuries and illnesses through running data analysis methods can help improve athletic performance. However, running data has the characteristics of multi-dimensional and long time series. It is difficult to process running data using traditional analysis methods, and the prediction accuracy is low. In response to this problem, this paper proposes an analysis method based on the Attention-LSTM model for injury prediction. In this model, LSTM uses the timing-related characteristics of the data for deep learning, which improves the long-term dependence of the RNN model, and adds an Attention mechanism to eliminate data redundancy and improve prediction accuracy. Take the athlete running training data set as an example, use the first 80% of the data set to train the model, and use the last 20% of the data set to test. The Attention-LSTM model and the RNN, LSTM, BiLSTM and Attention-BiLSTM models were used for experimental comparison. The results show that the MSE, RMSE, and MAE values predicted by the Attention-LSTM model are lower than the RNN and LSTM models, the R_Squred value is greater than the RNN and LSTM models, and the Attention-LSTM model calculation time is less than the BiLSTM and Attention-BiLSTM models. Comprehensively measuring the prediction effect and time cost factors, the Attention-LSTM model can effectively eliminate data redundancy, reduce prediction errors and improve prediction accuracy. It is suitable as a method for running data analysis.