c). Event:

22ND ANNUAL POPFEST CONFERENCE, LONDON, UK

Presentation title:

“Prospective Space-Time Scan Statistics (STSS) for Crime Prediction”

Abstract:

Allocating police resources proactively to areas of transiently elevated crime severity (emerging hotspots) is an effective strategy to reduce crime (Short et al. 2009). However, the effectiveness of this strategy depends on the capability of the hotspot detection method to provide adequate warnings (tipping points) and systematic monitoring of the emerging hotspots. The use of spacetime kernel density estimation (STKDE) as a hotspot detection method has previously been limited to visualisation of historical crime hotspots, which are then used to anticipate long-term police interventions. This approach is considered less useful for predictive policing in which short-term interventions are required in areas where a crime hotspot is rapidly emerging. This study considers the application of STKDE for prospective hotspot detection, in which emerging crime hotspots are identified before they reach their maximum level of severity. In our approach we segment historical crime datasets into small temporal partitions (daily and weekly), append the segmented dataset cumulatively while the STKDE of each cumulated dataset is being carried out. The preliminary results obtained show that STKDE has the potentials to give warnings and reveal emerging patterns in the crime dataset if implemented appropriately. This approach is currently being validated by comparison with other existing detection methods such as space-time scan statistics (STSS).

Speaker:

Monsuru Adepeju