L'identification des méthodes de classification de réseaux de neurones artificiels et des k plus proches voisins dans le dépistage des archers de haute performance à partir d'une sélection de paramètres de performance physique et motrice

Translated title of the contribution: The application of Artificial Neural Network and k-Nearest Neighbour classification models in the scouting of high-performance archers from a selected fitness and motor skill performance parameters

R. Muazu Musa*, A. P.P. Abdul Majeed, Z. Taha, M. R. Abdullah, A. B. Husin Musawi Maliki, N. Azura Kosni

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

43 Citations (Scopus)

Abstract

Objective: The utilization of artificial intelligence has been demonstrated in the literature to be effective for classification and prediction. Nevertheless, the application of k-Nearest Neighbour (k-NN) and Artificial Neural Network (ANN) specifically the conventional feed forward Multilayer Perceptron (MLP) model for forecasting and scouting of high-performance archers have not been fully utilized. The current investigation predicted high and low-performance archers from a set of selected fitness and motor skill parameters trained on two distinct machine learning algorithms viz. ANN and k-NN. Methods: A sample of 50 youth archers with the average age and standard deviation of (17.0 ± 0.56) recruited from varying youth archery schemes completed a one end archery score test. Standardize physical fitness and motor skill parameters measurements constituting of the hand grip, vertical jump, standing broad jump, static balance, upper muscle strength and the core muscle were carried out. The Hierarchical Agglomerative Cluster Analysis (HACA) with Mahalanobis’ distance was employed to group the archers with regard to the performance parameters assessed. The t statistic and Cohen's d effect size analysis were carried out on the group defined by the HACA to view through the performance differences of the archers. The ANN (single hidden layer with ten neurons) and k-NN (fine Euclidean-based) models were trained based on the measured performance variables. The tenfold cross-validation technique was utilized in the study. Results: The HACA grouped the archers into two distinct clusters namely; high-performance archers (HPA) and low-performance archers (LPA). It was observed from the t-statistic as well as the effect-size analysis that the performance of the HPA archers differed from the LPA in standing broad jump, hand grip, upper muscle strength as well as the archery shooting score P < 0.05 with a large to moderate effect-size d = 0.8–0.6. It was established that the ANN model outperformed the k-NN in the present study. The ANN demonstrated reasonably excellent classification on the evaluated indicators with a classification accuracy of 92% and a stronger Matthews correlation coefficient, i.e. 0.816 amongst other performance metrics in comparison to the k-NN model in classifying the HPA and the LPA. Conclusion: These findings are invaluable to coaches and sports officials, particularly in the identification of high-performance archers from a consolidation of the selected few evaluated fitness and motor skill performance parameters. As a consequence, this approach, in turn, would save resources, time and energy during a talent search program.

Translated title of the contributionThe application of Artificial Neural Network and k-Nearest Neighbour classification models in the scouting of high-performance archers from a selected fitness and motor skill performance parameters
Original languageFrench
Pages (from-to)e241-e249
JournalScience and Sports
Volume34
Issue number4
DOIs
Publication statusPublished - Sept 2019
Externally publishedYes

Keywords

  • Archery
  • Artificial intelligence
  • Fitness parameters
  • Motor skill
  • Talent scouting

Fingerprint

Dive into the research topics of 'The application of Artificial Neural Network and k-Nearest Neighbour classification models in the scouting of high-performance archers from a selected fitness and motor skill performance parameters'. Together they form a unique fingerprint.

Cite this