Cluster detection for spatio-temporal dengue cases at Selangor districts using multi-EigenSpot algorithm

Detecting clusters for spatio-temporal cases are becoming important to help hotspots detection for any seasonal outbreaks' cases such as dengue, covid-19, malaria etc. Cluster detection is classified into three types of clustering groups, which are spatial clustering, temporal clustering, and spatio-temporal clustering. In this study, spatio-temporal clustering is carried out to dengue datasets that were obtained from the Ministry of Health (MoH), Malaysia. Generally, the datasets were analyzed based on dengue cases reported for Selangor districts in years 2009 until 2013 to detect abnormal regions between the study areas. In health organization and epidemiology sectors, detection of cluster disease plays an important role to understand disease etiology and improve public health interventions strategy. Parametric assumptions commonly implemented in most of algorithm in cluster detections. However, the main limitation of the parametric assumptions are restrictions on the datasets' quality and type of clusters shapes. This study aims to detect the spatio-temporal clustering or hotspot regions of dengue cases for the districts of Selangor, Malaysia using a nonparametric algorithm (Multi-EigenSpot) to detect dengue clusters. The algorithm was deployed to the datasets using MATLAB software. This study has found that the most likely clusters were detected more efficiently when the algorithm removed the low-risk regions and low-risk time-point during scanning window search to avoid any false detection clusters. Different scope of clustering and geometric form of scanning window has significant contribution to the detected clusters. The finding in this study indicates that Petaling district is the most likely clusters which contributed the most of the reported dengue cases in Malaysia. © 2022 Author(s).