Publications

Generative models for spatial-temporal processes with applications to predictive criminology

Abstract

We present a generative model for spatialtemporal data that describes geographically distributed interactions between pairs of entities. We develop an efficient approximate algorithm to infer unknown participants in an event given the location and the time of the event. As a concrete application of the proposed approach, we focus on the problem of modeling inter–gang violence, where the objective is to infer the identities of participants in violent inter-gang attacks, based on the past observations of such attacks. We validate the model on synthetic as well as real–world data, and obtain very promising results on the identity–inference task. Furthermore, it is shown that combining both spatial and temporal information yields better accuracy than using either information separately.

Date
October 9, 2025
Authors
Y Cho, Aram Galstyan, Jeff Brantingham, George Tita