Bandwidth Exploration for Spatial-Temporal Kernel Density Visualizations

As an important and widely used data visualization tool for understanding datasets, spatial-temporal kernel density visualization (STKDV) has been extensively applied in a variety of domains, such as crime hotspot analysis, traffic accident hotspot analysis, and disease outbreak analysis. To obtain high-quality STKDV, it is typically necessary to adjust different spatial bandwidth values and temporal bandwidth values to generate multiple STKDVs, from which the optimal one is selected. However, the computational cost of a single STKDE is already prohibitively high, let alone multiple STKDVs. This substantial computational cost severely limits the bandwidth exploration for STKDV. To significantly improve the efficiency of the bandwidth exploration, we propose two optimization algorithms: a parallel voxel-based method named the PVSI Algorithm and a data-point-based method named the SPN Algorithm. Extensive comparison experiments have been performed on four real datasets, and the results demonstrated superior performance of the PVSI Algorithm.The PVSI Algorithm can be applied to bring near-real-time computing to the bandwidth exploration for STKDVs in the shortest time.