A1527
Title: A hybrid approach for the spatial clustering problem via the cross-entropy method
Authors: Nishanthi Raveendran - Western Sydney University (Australia) [presenting]
Georgy Sofronov - Macquarie University (Australia)
Abstract: Spatial data is often heterogeneous, meaning a single statistical model may not accurately describe the data. To address this, the data can be divided into several homogeneous regions or domains. The process of identifying these regions and their boundaries is known as spatial clustering (or segmentation) in spatial statistics. Spatial clustering has wide applications across various fields, including criminology, epidemiology, and ecology. The focus is on spatially correlated crime data. Due to the complexity of the model space, which is non-Euclidean after accounting for cluster label permutations, a hybrid approach combining the Cross-Entropy method with a genetic algorithm is proposed. The results demonstrate that the proposed algorithm effectively identifies homogeneous clusters within spatial binary data, providing valuable insights for studying the pattern of species distribution in ecology.
