Investigation of the Assessment Scales Applied to Adolescents With Psychiatric Disease by Using Smokeless Tobacco by Machine Learning Methods

Abstract

Abstract Purpose Adolescence is a fragile period in which all people live. This period can be more difficult for some people. In this difficult and fragile period, young people can suffer permanent psychological damage due to reasons such as social and family environment. One of these bad habits is smokeless tobacco. Unfortunately, the age of use may decrease worldwide due to reasons such as easy access and it can easily become addictive in adolescence, which is the sensitive period of human beings. Materials and Methods In our study, it was aimed to investigate the relationship between the use of smokeless tobacco and the use of machine learning methods in adolescents with psychiatric diagnoses. Various graded scale questions applied to adolescents were investigated with Embedded feature selection methods. Embedded methods; It can perform detailed feature selection calculations with three different calculations: Lasso, Gini and Permutation. Logistic Regression (LR) and Random Forest (RF) classification methods of selected features are used to select the most relevant features. Results Classification accuracy up to 0.98 (Lasso + LR = 0.98, Gini + RF = 0.95, Permutation + RF = 0.93) was calculated according to the selected features. Conclusion According to the results obtained from these classification methods, there is a goal of reaching the ideal by providing feedback. In our study, especially Lasso and Gini feature selections chose test questions related to smokeless tobacco use at a high rate. The permutation method also chose these test questions, but Lasso and Gini made relatively more choices.