Sistemas de classificação e auxílio ao diagnóstico de transtornos mentais em usuários de substâncias psicoativas com base em inteligência computacional

Abstract

Mental disorders are among the most prevalent diseases in the world and many studies have observed the relationship between the use of psychoactive substances with CMD or even depression, characterized by depressive, anxious and somatic symptoms, such as irritability, fatigue, insomnia and others. On the other hand, ML has been widely used to solve many problems in different areas. In this context, this study aims to test the effectiveness of ML as an auxiliary tool in the pre-diagnosis of CMD and depression, through the classification of users of psychoactive substances regarding the risk of depression and/or even CMD. The main objective is to obtain a model to predict the risk of depression and CMD, as well as to determine which factors contribute most to the prediction of the risk of depression and CMD. The databases used in this work were composed of 605 samples from people from eight cities in the state of Ceará, Brazil, collected from January to July 2019. The results showed the accuracy of the ML techniques tested in the prediction of CMD and depression , reaching an accuracy of 82.81% and 81.98% respectively, with emphasis on the Support Vector Machine (SVM) classifier with the Sequential Backward Selection (SBS) attribute selection technique in both databases. The results also showed that the use of tobacco derivatives, alcohol and cocaine/crack were the most significant factors in the classification in the databases, pointing out that the use of these psychoactive substances (SPA) caused the relapse, contributing to the individual’s return to the use of SPA, as well as which SPA were the most used. Thus, the study showed that the use of data mining (DM) and ML technique can significantly contribute to the pre-diagnosis of diseases such as mental disorders.