2026-05-04T19:54:08.233Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:11833 2026-02-03T17:33:45Z 3

Mental Disorder Classification via Modified Metaheuristic Optimized Machine Learning, Chapter in LNEE Lecture Notes in Electrical Engineering: PEIS 2025: Power Engineering and Intelligent Systems, Springer, volume 1459 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/11833 https://link.springer.com/chapter/10.1007/978-981-96-9716-8_11 L. Babic L. Jovanovic S. Kozakijevic M. Zivkovic T. Zivkovic S. Andjelic B. Radomirovic N. Bacanin Mental health conditions increasingly affect a significant portion of the population, posing serious challenges to public health systems worldwide. With growing awareness, accurate diagnosis and effective treatment play a vital role in providing care, supporting recovery, and maintaining individual well-being. Modern research has proposed various tests for detecting mental health conditions, leveraging advancements in technology and data analysis. However, several factors, including socioeconomic barriers and geographic limitations, can restrict individuals’ access to mental health professionals. This study explores the use of machine learning (ML), specifically categorical boosting (CatBoost), for detecting mental disorders based on reported indicators. Since the performance of classification algorithms heavily depends on proper hyperparameter selection, a modified metaheuristic algorithm is presented to address this challenge. This algorithm, based on particle swarm optimization (PSO), incorporates modern diversification control mechanisms to refine the optimization process further and enhance efficiency. The proposed optimizer is applied to manage hyperparameter tuning and improve classification accuracy in simulations conducted on a dataset available for free. The top-performing models showcase favorable results, effectively identifying conditions based on the input data provided, highlighting the potential of ML-driven solutions in mental health detection. bookPart Springer, Singapore 129 145 10.1007/978-981-96-9716-8_11 LNEE Lecture Notes in Electrical Engineering: PEIS 2025: Proceedings of International Conference on Power Engineering and Intelligent Systems, volume 1459