Modeling depression trends through continuous cognitive mapping

Modeling the global burden of depression presents significant challenges due to the chronic, dynamic, and heterogeneous nature of the disease. While Fuzzy Cognitive Maps (FCMs) provide an interpretable framework for modeling relational dynamics, conventional formulations often suffer from numerical saturation and a limited ability to capture structurally diverse solution spaces. In this paper, we propose a Continuous Fuzzy Cognitive Map (cFCM) framework to analyze longitudinal Disability Adjusted Life Years (DALYs) trajectories across 18 demographic cohorts from 1990 to 2021. The proposed model introduces a novel exponential activation function with a tunable nonlinearity parameter, ensuring numerical stability over long-term temporal horizons. To move beyond single-point representations, we employ a multimodal learning strategy, identifying multiple structurally distinct interaction patterns via k-means clustering across 100 independent training runs. Our findings uncover three stable network configurations, indicating that the global depression burden is governed by multiple latent demographic dependencies rather than a unique interaction structure. Centrality analysis identifies males aged 40–44 as the dominant hub, while a persistent sex asymmetry is observed, with male cohorts on average exhibiting a higher total degree connectivity compared to their female counterparts. Furthermore, polarity measures reveal a stable structural partition of the population, with one reinforcing and seventeen inhibitory drivers that remain consistent across the three learned configurations. These results demonstrate that the proposed cFCM framework effectively captures the robust, high-dimensional relational features of global mental health dynamics.