Balanced Random Forestを用いた10代青年の自殺念慮の予測

抄録

This study aimed to predict adolescents’ suicidal desires from their time perspective and family socioeconomic status (SES) in addition to theoretically related variables. Social survey data consisted of responses from 1,137 adolescents and 1,906 parents, which were analyzed using the machine learning method Balanced Random Forest. The results showed that all participants who had suicidal desires were predicted as having suicidal desires, and the overall accuracy was about 80 percent. Congruence of adolescents’ and their parents’ time perspectives with respect to the adolescents’ educational stage and family SES were high on both indices of importance, among all variables used in this study that were considered important based on previous findings and theories. Therefore, the importance of congruent adolescent-parent time perspectives and family SES for the prediction of suicidal ideation among Japanese teenagers was demonstrated. Therefore, prevention specific to this developmental stage is necessary, keeping the aforementioned factors in mind.