Predictive analysis of dominant hand grip strength among young children aged 6-15 years using machine learning techniques: a decision tree and regression analysis.

Background: This study aimed to investigate and understand predictor variables and isolate the exact roles of anthropometric and demographic variables in the hand grip strength of young children. Material and methods: In total, 315 male and female children participated in the study and 11 participants were excluded, therefore, 304 participants completed the assessments. Anthropometric measurements were collected at the time of study, along with age, height, weight, circumference of the hand, hand span, hand length, palm length, and hand grip strength (HGS) was measured. Both decision tree and regression machine learning analyses were used to isolate the relative contribution of independent features in predicting the targeted grip strength of children. Results: Two predictive models were developed to understand the role of predictor variables in dominant hand HGS for both boys and girls. For boys, the decision tree was found to be the best model with the lowest error in predicting HGS. The respondents' age, hand span, and weight were the most significant contributors to male hand grip strength. For the boys under 9.5 years of age, based on the decision tree analysis, weight (split at 27.5 kg) was found to be the most significant predictor. Furthermore, for the boys under 14.5 years of age, weight (split at 46.7 kg) remained the most important predictor. For boys 14.5 years and older, hand span was important in predicting handgrip strength. Backward regression was found to be the best model for predicting female hand grip strength. The Conclusion: Anthropometric factors played a significant role in hand grip strength. Age, weight, and a larger hand span were found to be significant in impacting male HGS, while BMI, hand length, and palm length contributed to higher grip strength among the girls.