Crime Prediction by Data-Driven Green's Function method
share
We present an algorithm for crime prediction based on the near-repeat victimization model solved by a Green's function scheme. The Green's function is generated from spatio-temporal correlations of a density of crime events in a historical dataset. We examine the accuracy of our method by applying it to the open data of burglaries in Chicago and New York City. We find that the cascade of the crimes has a long-time, logarithmic tail, which is consistent with an earlier study on other data of burglaries. The presented method is a powerful tool not only to predict crimes but also to analyze their correlations, because the Green's function can describe how a past crime influences the future events.
READ FULL TEXT