<![CDATA[Background: Current methods for managing Dengue Hemorrhagic Fever (DHF) in East Java often separate temporal and spatial analyses, hindering proactive public health interventions. To address this limitation, this study develops an integrated spatio-temporal framework that combines multi-model forecasting and geospatial analysis to create localized dengue projections and risk maps. Subjects and Method: An ecological spatio-temporal design was conducted across 38 districts/ cities in East Java using secondary data from 2013–2023. The dependent variable was annual dengue incidence, while independent variables comprised natural factors (e.g., rainfall, humidity) and social factors (e.g., population density, poverty). Data were analyzed using Geographically Weighted Regression (GWR) for spatial modeling. Time series forecasting for the 2024–2028 period was conducted using Trigonometric seasonality, Box-Cox transformation (TBATS), Autoregressive Integrated Moving Average (ARIMA), Prophet, and Neural NETwork AutoRegression (NNETAR) models, evaluated by Root Mean Square Error (RMSE). Results: Rainfall demonstrated the strongest natural association (R²=0.49; p=0.022), followed by humidity (R²=0.26; p<0.001). Epidemiologically, population density emerged as the key social determinant (R²=0.50; p=0.005). TBATS achieved the highest forecasting accuracy with the lowest RMSE (47.657), outperforming ARIMA (202.980), Prophet (172.809), and NNETAR (157.653). Spatial risk mapping revealed that high-risk social clusters are heavily concentrated in the eastern 'Tapal Kuda' region and northern coastal metropolitan corridors. Conclusion: Dengue incidence in East Java is significantly driven by climatic factors and spatially varying social vulnerabilities. The integration of TBATS forecasting and GWR spatial modeling provides a robust framework for anticipating future case burdens and identifying priority intervention areas, thereby supporting highly targeted public health planning.]]>
