Identifying the most important predictors of different muscle groups using electromyography, regression models and Artificial Neural Networks in the flat bench press

doi: 10.29359/JOHPAH.1.4.02

Identifying the most important predictors of different muscle groups using electromyography, regression models and Artificial Neural Networks in the flat bench press

Wojciech Smółka1, Jan Pilch2, Magdalena Krawczyk3, Małgorzata Kowalczyk3, Małgorzata Bąk3, Angelina Ignatjeva3
1 Clinical Department of Laryngology, School of Medicine in Katowice Medical University of Silesia, Katowice, Poland
2 Department of Physiological and Medical Sciences The Jerzy Kukuczka Academy of Physical Education, Katowice, Poland
3 Institute of Sport Sciences The Jerzy Kukuczka Academy of Physical Education, Katowice, Polan

abstract

Background: ‪‪The main objective of this study was to determine the input of different muscle groups during the flat bench press with different external loads and to determine whether regression models or Artificial Neural Networks (ANN) models predict sports results more precisely and indirectly better support and optimize the athletes’ selection process in the particular strength exercises.

Material and methods: The activity of four muscles was measured in four tasks: the pectoralis major (PM), the anterior deltoid (AD), the lateral head of triceps brachii (TB), and the latissimus dorsi (LD).

Results: The greatest increase in bioelectrical activity with increased external loads was observed on the LD during the descending phase of movement. Then, on the basis of results of 51 athletes, mathematical models were created and an additional study was conducted with the experimental group in order to verify the previously created models which were based on one group of 15 athletes. The regression models and perceptron networks demonstrated their capacity for making generalization and predicting sports results.

Conclusions: ‪The results of the investigation show that the created neural models (9-4-1 structure) offer much higher quality of prediction than a nonlinear regression model.

Key words: non-linear models, ANN, sports results, flat bench press.