Genetic Algorithm-based Feature Selection for Depression Scale Prediction

This study aimed to improve the performance of machine learning prediction through feature selection using a genetic algorithm (GA) for predicting depression of elderly people based on the survey data of Korean Longitudinal Study of Aging (KLoSA) performed in South Korea. The proposed feature selection method finds an optimized feature set through a fitness function design that maximizes the correlations between the features selected and the label to be predicted while minimizing the correlations between the selected features by using GA. The effectiveness of the proposed GA was shown through comparative experiments.