e). Event:

THE 26th ANNUAL CONFER. OF THE INTL. ENVIR. SOCIETY, EDINBURGH, SCOTLAND, UK

Presentation title:

“A Spatially-Integrated Sparse Point Time Series (SSPT) Prediction Method”

Abstract:

The distribution of many physical and sociological processes such as crime and disease, is inherently prone to high zero counts. That said, the predicting hotspot modelling of this distribution becomes difficult due to the lack of methodologies to adequately capture the excess zero counts. In this research, we developed a new predictive hotspot method that is robust to excess zeros of sparse spatio-temporal point process (SSPT). The method is implemented for prediction on arbitrary square grids and on street network. We used two crime datasets to evaluate the new method: Camden Borough of London (United Kingdom) and South-side Chicago (United States). The preliminary results show that both the grid cell-based and street network-based predictions show significantly improved accuracy over the existing methods. Also, a comparison of the grid cell-based and street network-based versions of our newly developed method revealed street network to be a better spatial unit for predictive hotspot in comparison to grid spatial unit.

Speaker:

Monsuru Adepeju