The prevention and prediction of suicide and depression: a population-based study on machine learning approach