L.A. Kovalkova1, A.A. Kochkarov2, E.Yu. Shchetinin3

1–3 Financial University under the Government of the Russian Federation (Moscow, Russia)

1 mila.kovalkova.ai@gmail.com, 2 akochkarov@fa.ru, 3 eyshchetinin@fa.ru

This text presents a comprehensive analysis of the application of both classical and modern methods for analyzing psychometric data to assess individuals' mental states. Regression and classification are the main classes of machine learning methods used in this study. The task involves predicting a dependent variable representing levels of depression, anxiety, or stress on a scale of 0 to 4. Various machine learning models, including random forest, linear regression, gradient boosting, and extreme gradient boosting, are employed. The original dataset, which includes demographic data and features of the dispositional personality model, is divided into training and testing sets. Performance evaluation of regression algorithms is conducted using metrics such as mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R2). Results underscore the inadequacy of regression models for this specific task, prompting a transition to classification methods. Subsequently, classification methods like extreme gradient boosting are applied, utilizing techniques such as Synthetic Minority Over-sampling Technique (SMOTE) for class balancing. Metrics like recall, precision, and F1-score are used for evaluation. The study highlights the superiority of classification methods, particularly the extreme gradient boosting algorithm, in predicting mental health status. Designing independent variables demonstrates that significant reduction in the number of parameters minimally diminishes prediction accuracy. Verification of the extreme gradient boosting algorithm on a separate dataset confirms the necessity for data integrity and homogeneity concerning the target group in light of latent interdependencies among variables. In conclusion, the study high-lights the potential of machine learning methods in assessing mental health and underscores the prospects of incorporating psychometric data in conjunction with physiological data for more accurate prediction of mental health status.

Kovalkova L.A., Kochkarov A.A., Shchetinin E.Yu. Application of machine learning methods for analysis of the mental state of a person. Neurocomputers. 2024. V. 26. № 2. Р. 49-58. DOI: https://doi.org/10.18127/ j19998554-202402-05 (In Russian)

1. Anishchenko L.N., Turk A.V., Bugaev A.S. Machine learning in psycho-physiological state assessment by bioradar means. Biomedical radioelectronics. 2020. V. 23. № 4. P. 31–40. DOI 10.18127/j15604136-2004-05. (in Russian)
2. Myshenkov K.S, Nekoula Haddad Neurological disease prediction using machine learning methods. Dynamics of complex systems. 2021. V. 16. № 1. P. 66−74. DOI 10.18127/j19997493-202201-07. (In Russian)
3. Alkhanov A.A. Machine learning and its application in the modern world. Problems of science. 2021. № 7(66). P. 25–27. (In Russian)
4. Poryaeva E.P., Evstafyeva V.A. Artificial intelligence in medicine. Bulletin of Science and education. 2019. № 6-2(60). P. 15–18. (In Russian)
5. Nevenchanny S., Cherginets S. Psychological conditions of military personnel in combat conditions. Ways of forming the moral readiness and psychological ability of military personnel for combat operations. Army Collection. 2020. № 8. P. 100–106. (In Russian)
6. Open-source psychometric project. [Electronic resource] – Access mode: http://openpsychometrics.org/, date of reference 31.08.2023.
7. A large psychological dictionary. Ed.by B.G. Meshcheryakov, V.P. Zinchenko. Ed. 4th. SPB: Prime-Euroznak. 2009. 811 p. ISBN 978-5-93878-663-9. (In Russian)
8. The University of New South Wales, School of Psychology. [Electronic resource] – Access mode: http://www2.psy.unsw.edu.au/dass/, date of reference 31.08.2023.
9. Khromov A.N. Five-factor personality questionnaire: Study method. stipend. Kurgan: Kurgan State University. 2000. 23 p. (In Russian)
10. Muller A. An introduction to machine learning using Python. A guide for data specialists. M.: Alpha-book. 2017. 697 p. (In Russian)
11. Gholamy A., Kreinovich V., Kosheleva O. Why 70/30 or 80/20 Relation Between Training and Testing Sets: A Pedagogical Explanation. Journal of Intelligent Technologies and Applied Statistics. 2018. V. 11. № 2. P. 105–111.
12. Torgo L., Gama J. Regression using classification algorithms. Intelligent Data Analysis. 1997. V. 1. № 4. P. 275–292. DOI 10.3233/ida-1997-1405.
13. Pedro D. The Supreme algorithm: how machine learning will change our world. Moscow: Gostekhizdat. 2015. 989 p. (In Russian)
14. Kruglov V.V., Borisov V.V. Artificial neural networks. Theory and practice. Ed. 2th. M.: Hotline-Telecom. 2002. 382 p. (In Russian)
15. Azure Machine Learning Documentation. [Electronic resource] – Access mode: https://learn.microsoft.com/ru-ru/azure/machine-learning/component-reference/smote?view=azureml-api-2, date of reference 21.08.2023. (In Russian)
16. Chawla N.V., Bowyer K.W., Hall L.O., Kegelmeyer W.P. SMOTE: synthetic minority over-sampling technique. Journal of Artificial Intelligence Research. 2002. V. 16. P. 321–357. DOI 10.1613/jair.953/
17. Singh A. Machine learning with Python. 2019. 137 p.
18. Chris E. Machine learning using Python. Collection of recipes. St. Petersburg: BHV-Petersburg. 2019. 384 p. (In Russian)