NEURAL NETWORKS IN SPORTS MODELING

Abstract

In the field of sports science, artificial intelligence and machine learning have achieved significant success in recent years. The experience of best practice indicates the effectiveness of using artificial neural networks (ANN) in modeling sports technique. However, researches on the use of ANN in sports are presented fragmentarily and mainly by English-speaking authors. The research objective is to analyze, systematize and generalize the areas of application of artificial neural networks as a modeling tool in sports.

Methodology: systematic content analysis of scientific and methodological literature and Internet data.

Scientific novelty: the use of neural networks for modeling in sports is systematized according to scientific areas: modeling of athlete's sports fitness indicators, sports result forecasting, media images on athlete action recognition, sports selection, injury prevention. The main advantages of neural networks: 1) they are able to approximate any continuous function, and thus the researcher does not need to have any hypotheses regarding the basic model; 2) the connection between the input and output is determined in the process of training the neural network.

Conclusions. The use of neural network technology in sports disciplines allows to predict sports results, and also significantly increases the effectiveness of the training process due to an individual approach, accurate biomechanical assessment of the technique of the main motor action and making an objectively justified rational decision on its correction.

Today, two statistical modeling methods are mainly used in the field of sports – regression analysis and neural networks. Unlike regression equations, the modeling method using neural networks is non-parametric and therefore more flexible. The model based on a neural network does not require prior definition of mathematical expressions of the prediction model and has higher prediction accuracy.