Predicting Performance in Women Cricketers through Anthropometric Indicators; Using Machine Learning Approach

Abstract

The success of cricket players is affected by physical and technical factors. However, little research is available on the prediction of the performance of female cricketers through their anthropometry and new analysis techniques in Pakistan. This study was an effort to use machine learning to predict cricket performance in the female cricket players from 17-21 years old and belonging to six areas in Khyber Pakhtunkhwa (KP). A total of 248 players were selected by sampling in different stages viz districts, colleges and randomly selecting teams. The anthropometric variables were height, weight, BMI, fat distribution, skin fold thickness, arm length and leg length. Cricket skills and fitness scores were taken as performance outcomes. Among the machine learning techniques, Random Forest regression was applied to improve the prediction accuracy as well as to find out the key performance determinants. The results showed that limb lengths, bowling speed and throwing accuracy were the most important factors in fitness and cricket performance and the body fat measures had lesser effect. The research demonstrated that machine learning supports performance prediction, facilitating evidence-based decision-making in the talent identification, training choice, and decision making in women's cricket development in Pakistan.

Keywords: Machine learning, Anthropometry, women's cricket, performance prediction, random forest.