A Hybrid Analytical Framework for Exploring Dengue Incidence and Climatic Correlations in Peninsular Malaysia Using ARDL-SVM and Spatial Clustering

This study presents a hybrid analytical framework designed to investigate the complex relationships between climatic variables and dengue incidence across districts in Peninsular Malaysia. By integrating Autoregressive Distributed Lag (ARDL) models with Support Vector Machine (SVM) techniques, the framework leverages both traditional econometric modeling and machine learning to enhance the predictive performance of dengue outbreaks based on temperature, humidity, and rainfall.The research utilizes historical time-series data (2013–2020) for model training, with 2021 data reserved for validation. The ARDL component identifies the statistically significant lagged effects of climatic variables on dengue cases, while the SVM refines the nonlinear patterns and improves forecasting accuracy. To assess spatial variation, Local Indicators of Spatial Association (LISA) are employed to detect spatial clusters and outliers in both observed and forecasted dengue incidence values.Furthermore, the study includes an innovative weighting approach to quantify the impact of climatic factors at the district level by combining ARDL-derived coefficients with the standard deviation of climate and dengue variables. These impact values are then categorized into three levels (low, medium, and high), providing a spatially explicit understanding of climate sensitivity on dengue transmission.The results highlight significant spatial heterogeneity, with urban districts generally exhibiting higher root mean square error (RMSE) values due to population density and intervention complexities, while rural districts showed more climate-driven transmission patterns. The integrated ARDL-SVM-LISA framework demonstrates its potential as a robust tool for early warning systems, targeted interventions, and public health policy planning.