طاهره اسماعیلی ابهریان

The purpose of this paper was to analyze the performance of elite taekwondo athletes using machine learning approaches to achieve three objectives: categorizing performance into four clusters ranging from excellent to poor, identifying key physical characteristics influencing performance, and predicting medal-winning potential in world competitions. This study employs the National Olympic Academy dataset of Iranian taekwondo athletes’ physical fitness and anthropometric records from 1996 to 2019 to develop descriptive and predictive models. In datasets comprising 999 female and 1560 male records, SOM-means and SOM-spectral clustering algorithms achieved average test efficiencies of 80%, for Silhouette, and 20% for Davies-Bouldin in the female dataset, and 79% and 34% respectively, in the male dataset, identifying four performance clusters based on physical attributes and medal distribution. A semi-supervised learning model with the CPLE-Learning algorithm demonstrated medal prediction capabilities, with accuracy rates of 68%, 59%, and 73% for predicting gold, silver, and bronze medals in the female dataset, and 58%, 61%, and 54% in the male dataset. These findings highlight the effectiveness of machine learning in sports performance analysis, offering valuable insights for managing taekwondo athletes and enhancing physical preparation strategies.