Cluster analysis and artificial neural networks to assess and diagnosis suicide ideation in school adolescents

Abstract

Worldwide, suicide is considered a health problem where adolescents are most at risk. This work developed a method to assess and predict an early diagnosis of suicidal ideation among school adolescents through multivariate techniques: cluster analysis and artificial neural networks. Variables related to suicidal thoughts, plans and manifestation were analyzed in (n=638) adolescents. Cluster analysis identified 73.2% of adolescents with low suicidal ideation, 18.5% with medium suicidal ideation and 8.3% with high suicidal ideation. A neural network was designed with a correct classification capacity of 95.5%. The proposed method can discriminate and diagnose suicidal ideation in school adolescents. These results seek to create and develop initiatives focused on early detection and intervention to implementing educational and public policies preventing suicide among adolescents.