Research on Training Quantity and Sports Performance Based on LSTM Model: A Case Study of Marathon Sports

With the increasing awareness of national health, the number of marathon events in China is constantly increasing, and the number of amateur marathon enthusiasts is also growing exponentially. However, they have not received professional training and have a certain pursuit of competition results. In daily training, they are prone to low training efficiency due to unreasonable training plans. This article mainly analyzes the relationship between training volume and sports performance by establishing a mathematical model, in order to develop efficient training plans for marathon enthusiasts and help them achieve better competition results. This article constructs a stationary time series model by visualizing and analyzing some training volume time series graphs, setting LSTM neural network parameters, and using Python for rolling prediction. Some prediction graphs reflect an increasing trend in sports performance in the next month. Through analysis, we conclude that increasing training volume has a promoting effect on improving sports performance, but training volume is not absolutely positively correlated with sports performance. As the amount of training increases, the extent of performance improvement will gradually decrease, and there may even be a plateau period or a decline in performance. Therefore, in the actual training process, attention should be paid to physical load and fatigue index to develop a reasonable training plan in order to achieve maximum training benefits. Finally, error analysis, sensitivity analysis of historical data volume, testing, and improvement were conducted on the ARIMA model based on the above established models.