Depression Emotion Multi-Label Classification Using Everytime Platform With DSM-5 Diagnostic Criteria

• Published in: IEEE access Vol. 11; pp. 89093 - 89106
• Main Authors: Park, Dabin, Lim, Semin, Choi, Yurim, Oh, Hayoung
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: AI in psychiatry; Artificial intelligence; Blogs; Classification; Criteria; Data models; Depression; Diagnostic systems; Dictionaries; DSM-5; gated recurrent unit; Hidden Markov models; KoBERT; Labeling; Labels; Mental depression; Mental disorders; multi-label classification; Natural language processing; Predictive models; Psychiatry; Social networking (online); Statistical analysis; Young adults
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2023.3305477

The prevalence of depressive disorder in South Korea is the highest among countries in the Organization for Economic Cooperation and Development. The rate of depressive disorders among young people in their 20s is significantly high. Because depressive disorders lead to a decrease in daily function, it is important to identify them quickly and treat them appropriately. In many countries, including Korea, mental disorders are diagnosed using the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) classification criteria for mental disorders. However, research on the prediction of depressive disorders using DSM-5 classification criteria is sparse. In this study, we designed a method of predicting depressive disorders by identifying the characteristics of people in their twenties who are experiencing depression. We employed data from the Everytime platform and created a multi-label classification methodology that meets the DSM-5 diagnostic criteria for depressive disorders. Moreover, a gated recurrent unit was used to calculate the probability for each label.