Genetic Algorithm-based Feature Selection for Machine Learning System Diagnosing Sarcopenia

Abstract

This study investigates whether a genetic algorithm (GA) can improve the performance of machine learning for sarcopenia diagnosis. An essential aspect of applying feature selection to machine learning for diagnosing sarcopenia is the selection of features that directly affect the diagnosis. To determine whether the GA can perform this logic effectively, we performed feature selection using the Korean Longitudinal Study of Aging (KLoSA) survey data. This study is significant because it implements feature selection using GA and shows that diagnosis performance is improved compared to other machine learning methods without feature selection. In addition, the results showed that GAs could improve the diagnosis of sarcopenia in future research.