Satpura: A Novel Framework for Density Estimation, Hotspot Discovery, Change Analysis, and Change-based Alerts
MetadataShow full item record
Due to the technological advancement in remote sensors and sensor networks, different types of spatio-temporal data are increasingly available. Spatio-temporal data analysis has applications in many fields, including criminology, epidemiology, and traffic analysis. The main focus of this research is to develop a generic analysis framework called Satpura, which provides density estimation, hotspot discovery, and change analysis capabilities for spatial data. The framework supports naïve, and kernel density estimation approaches for raw and relative densities. To identify density hotspots, we designed a novel hotspot discovery technique that generates rectangular hotspots for a given density threshold. We also developed a post-processing technique to remove redundant and highly overlapping hotspots. Since the density threshold plays a significant role in hotspot generation, we developed an automatic density threshold selection technique. Additionally, we developed evaluation metrics to assess the quality of the hotspots. To address change analysis, we developed two techniques: density-based change analysis, which is used to find the regions where there is a high density change with time, and hotspot-density-based change analysis, which is used to identify the density changes that occur in hotspots over time. Based on the change analysis, density-change-based alerts and hotspot-density-change-based alerts are provided by Satpura. Satpura, which was developed in Python as a web-based application, was used to analyze an Austin crime dataset. It successfully identified crime hotspots, and it analyzed changes that occurred in criminal activity. Then, an alert system was implemented to warn the public of new crime hotspots. Satpura was also used to analyze an Austin traffic accident dataset.