Spatial-Temporal-Textual Point Processes for Crime Linkage Detection
This addresses crime linkage detection for law enforcement, but it is incremental as it builds on existing Hawkes processes by incorporating text embeddings.
The paper tackled the problem of detecting linkages between crime incidents using limited information like text, time, and location, by proposing a spatio-temporal-textual point process framework, and demonstrated improved performance over state-of-the-art methods in real data.
Crimes emerge out of complex interactions of human behaviors and situations. Linkages between crime incidents are highly complex. Detecting crime linkage given a set of incidents is a highly challenging task since we only have limited information, including text descriptions, incident times, and locations. In practice, there are very few labels. We propose a new statistical modeling framework for {\it spatio-temporal-textual} data and demonstrate its usage on crime linkage detection. We capture linkages of crime incidents via multivariate marked spatio-temporal Hawkes processes and treat embedding vectors of the free-text as {\it marks} of the incident, inspired by the notion of {\it modus operandi} (M.O.) in crime analysis. Numerical results using real data demonstrate the good performance of our method as well as reveals interesting patterns in the crime data: the joint modeling of space, time, and text information enhances crime linkage detection compared with the state-of-the-art, and the learned spatial dependence from data can be useful for police operations.