Predictors of depression in middle adulthood: A longitudinal machine learning model

This study aimed to predict the risk of depression, influencing factors, and gender differences in middle adulthood through applying machine learning models. We selected 2,674 middle-adult-aged participants from the China Family Panel Studies and used a combination of long short-term memory and machine learning models for prediction. Combining long short-term memory modeling with machine learning models significantly enhanced depression prediction among individuals in middle adulthood. Among the six models we examined, the eXtreme Gradient Boosting model performed the best. Further analysis of influencing factors revealed that happiness, self-rated health, and awareness of social issues were the most impactful factors in predicting the risk of depression. Further, the influencing factors varied between genders: for men, happiness, frequency of physical exercise, and job-income satisfaction were paramount, while for women, happiness, job-promotion satisfaction, and self-rated health were the key factors. Implications of the findings are discussed for theory and practice.