Exploratory topological data analysis for spatio-temporal knowledge discovery in epidemiology

Dengue remains a critical public health concern in Malaysia, exerting significant pressure on the healthcare system and impacting socio-economic stability. Effective control strategies require accurate interpretation of complex spatio-temporal epidemiological patterns, which is often challenging using conventional approaches. To address this, the present study proposes a topological-based framework for exploratory analysis of dengue dynamics across Malaysian districts using the Mapper algorithm. The Mapper graphs, analyzed using network metrics such as component density (0.158-1.0), path structures, and negative assortativity (down to -0.71), reveal distinct spatial severity regimes. Large, weakly connected components correspond to low-severity areas, while smaller, denser, or isolated components indicate moderate-to-high severity regions, highlighting how network topology captures spatial differences in dengue transmission. Levels of modularity (up to 0.36) further indicate a decomposition of districts into structurally distinct variability groups. Geographical and spatio-temporal examinations of representative Mapper components demonstrate the ability of the proposed approach to filter noise, preserve meaningful temporal structure, and differentiate coherent seasonal behavior from anomalous dynamics. These findings highlight the potential of topological-based methods to enhance exploratory epidemiological analysis and support data-driven understanding of heterogeneous dengue patterns at the district level.