Neural networks for forecasting and modeling training in track-and-field athletics

Aim. The article deals with the application of neural networks for forecasting the most optimal ways and intensity of training for 400-meter runners at the stage of performance improvement. Materials and methods. 400-meter male runners aged 18-21 participated in the study (8 runners of the first rank, 3 runners with the rank of Candidate for Master of Sport). During the study, we used Neural Network v2.4.2 software developed by Jwsoft.Net. Initial data consisted of 8 indicators for each athlete (n = 10) taken in compliance with the months of a one-year training cycle 2014/2015, 2015/2016, 2016/2017. Network training was performed with the algorithm of the error back propagation. Results. To simulate physical preparedness of 400-meter runners in 2016/2017, we inserted into a trained neural network the parameters of monthly volume load, which allowed us to forecast competition results for the 400-meter runners of the first rank and of the rank of Candidate for Master of Sport. The reliability of forecasting is 98-99 %. The method proposed based on the application of the neural network allows to quickly estimate the dynamics of physical preparedness. This provides the reliability and quality of forecasting based on the training plan. Conclusion. The application of neural networks will allow to determine the most optimal ways and volumes of training. The coach will have a tool, which allows him to make effective decisions about the correction of training.