Determining the optimal spatial scan extent (K) of a Prospective space-time scan statistics (PSTSS) that maximises the predictive accuracy of crime prediction

The input value for the maximum spatial scan extent (K) of a prospective space-time scan statistical (PSTSS) technique is one important parameter that ensures accurate detection of the predictive hotspot of geographical point events. Currently, there is no general consensus on how to determine the optimal value of K that maximises the predictive accuracy, especially in crime hotspot prediction. To address this issue, this study proposes a strategy in which different values of K are used to generate accuracy profiles that can be compared to determine the optimal value of K. A case study presented shows that the best accuracy is obtained at K=250m, which is equivalent to the spatial aggregation of the crime dataset used. It is contended that this strategy could be extended to other geographical surveillance tasks.