Applying random forest method to analyze elderly fitness training routine data

Abstract

This study used the random forest algorithm to predict Senior Fitness test results on the execution of Synchronized Monitoring Analysis Record Care (SMARC) programs with the aim of aiding healthcare professionals in modifying patients’ training routines to improve their effectiveness. Twenty-three subjects in a community center performed a fitness training routine using the SMARC series of equipment and training modes, and took timed “Up and Go” tests before and after their performances. The 74 combined features (categorical + numerical) of the series were used as input features, and performance was measured by the Timed Up and Go (TUG) score. The results show that the top five features ranked with the highest importance were associated with Machines F (16.5%), D (15.4%), E (13.9%), H (13.9%), and B (12%), with 35% unassignable. The results can aid healthcare professionals in planning and adjusting more targeted health-promotion exercises programs using assistive devices for the elderly.