International Conference on Information and Knowledge Technology

Home / دوازدهمین کنفرانس بین المللی فناوری اطلاعات و دانش

Predicting Suicide Risk in Adolescents with Random Forest for Unbalanced Data Management

Authors :

Fatemeh Rabbani¹ Behrooz Masoumi² Mohammad Reza Keyvanpour³

1- دانشگاه آزاد اسلامی واحد قزوین 2- دانشگاه آزاد اسلامی واحد قزوین 3- دانشگاه الزهرا(س)

Keywords :

Suicide risk, Random forest, unbalanced data, Classification

Abstract :

Suicide is one of the major concerns of public health. Studies indicate the increasing prevalence of suicide, especially among adolescents. The risk factors of suicide include biological, psychological, clinical, social, and environmental factors. Involvement of various risk factors in suicide means that suicide risk in an individual is challenging; thus, to identify high-risk groups in public, a suicide risk prediction model is necessary. Today, employing machine learning and classification methods are widely used to predict suicide risk. One of the challenges of this context is unbalanced data that affect the efficiency of the prediction model. In this paper, two sampling methods are proposed to improve the performance of classifying unbalanced data, aiming to evaluate suicide risk in adolescents. In the proposed method, after balancing the dataset using sampling methods, the data is classified using random forest. The results show that the total accuracy of predicting suicide in adolescents is 0.99, with a sensitivity of 1 and specificity of 0.98. Therefore, the random forest model can predict suicide risk with high accuracy.