Travma Sonrası Stres Bozukluğu Yönetimi: Makine Öğrenmesi Yaklaşımı

Abstract

Post-traumatic stress disorder (PTSD) is defined as a traumatic injury developed after facing or witnessing a life-threatening experience or event such as a natural disaster, a pandemic, a serious accident, a terrorist act, war/combat, rape or other violent personal assault. Machine Learning (ML) has been widening its scope on psychological and physical healthcare for a decade by predicting detecting, personalizing, digitalizing, preventing risks, monitoring, and classifying PTSD based clinical mental diseases. In this study, we predict PTSD scores of the participants obtained from Mississippi-Civilian Version Scale and DSM-5 (PCL-5) Scale by applying ML. For our experiments we used the following methods namely k-nearest neighbor (k-nn), support vector machine (SVM), decision tree (DT), Gaussian Naive Bayes (GNB) and artificial neural networks (ANN). According to the comparison of the prediction results Considering PTSD prediction classification performance results for Mississippi (Civilian version) scale data set in comparison to the above mentioned methods, ANN offers the best prediction in terms of accuracy, F1 score and recall. However, Gaussian Naive Bayes (GNB) gives the best prediction score in terms of precision. On the other hand, when we applied all these methods to DSM-5 (PCL-5) scale data set, we have observed that ANN offers the best prediction in terms of accuracies, F1 score and precision. Nevertheless, in terms of recall Gaussian Naive Bayes (GNB) gives the best prediction score. By applying all the methods to these two different data sets and comparing the results, we demonstrate which method can be more efficient in prediction, diagnosis and monitoring the patients with PTSD.