Using Permutation Entropy to Evaluate Spatial Predictability in Urban Crime

Predicting where the crime occurs is essential and significant for preventive policing which is an important action for economic benefits, urban building and human safety. Predictability is a theoretical bound for the prediction performance in human behaviour based on limited data. Current approaches to predictability are usually based on human mobility, with the development of electronic information systems in the police system, more datasets about urban crime can be obtained. Therefore, it is possible to study the predictability of urban crime. To address this, this study uses permutation entropy as the measure to evaluate the spatial predictability of urban crime. The method has been evaluated using urban cities’ public crime datasets (Washington DC, Denver, New York, and Vancouver) from years 2010–2022. The results prove the hypothesis of correlation between the space scales of data and the level of spatial predictability, which can guide the prediction of the urban crime algorithms.