Machine learning applied to the diagnosis of depressive and bipolar disorders: a study with real patients

• Larissa Mitie Curi Hirai UFT
• Alexandre Tadeu Rossini da Silva UFT

DOI: https://doi.org/10.5753/sbsc_estendido.2025.8612

Abstract

This paper presents a comparison between three machine learning algorithms applied to the diagnosis of mental disorders (depression, bipolar type I and II) based on data from 120 real patients. The methods used were random forest, KNN, and neural network, evaluated using accuracy, confusion matrix, and AUC metrics. The neural network with OneHotEncoder encoding showed the best performance, with accuracy above 95%, highlighting the potential of using AI to support the clinical diagnosis of mental diseases.

References

APA, A. P. A. (2014). DSM-5: Manual diagnóstico e estatístico de transtornos mentais. Artmed Editora.

Dattani, S., Rodés-Guirao, L., Ritchie, H., and Roser, M. (2023). Mental health. Our World in Data. [link].

IHME (2024). Global burden of disease. Disponível em: [link]. Acessado em: 5 mai. 2024.

Karbalaeipour, H., Damari, S., Zolfagharnasab, M. H., and Haghdadi, A. (2023). A collection of 120 psychology patients with 17 essential symptoms to diagnose mania bipolar disorder, depressive bipolar disorder, major depressive disorder, and normal individuals.

OMS, O. (2023a). Depressive disorder (depression). Disponível em: [link]. Acessado em: 09 jun. 2024.

OMS, O. (2023b). International Classification of Diseases, 11th Revision (ICD-11). Retrieved from [link].

Shearer, C. (2000). The crisp-dm model: the new blueprint for data mining. Journal of data warehousing, 5(4):13–22.