Event scene modeling method based on dynamic crime network

Abstract

The invention discloses an event scene modeling method based on a dynamic crime network. The method comprises the following steps: firstly, processing a spatial feature group in a crime event; using a utility function to conclude personal information contained in criminal network nodes to analysis of a criminal network, and abstracting a criminal activity into a model with maximized transaction benefits; and finally, solving an optimal solution under a constraint condition. According to the invention, the dynamic interaction between the criminal event and different crime types and crime sites is ensured, and the variability of the criminal event in a dynamically changing crime scene is protected.

Description

technical field [0001] The invention belongs to the technical field of criminology, and in particular relates to an event scene modeling method. Background technique [0002] The existing scene modeling methods for crime events can be summarized into two main ways. The first method is to abstract and model crime events from time and space feature dimensions based on spatio-temporal crime hotspot data. These works can use GIS technology to obtain time series models, or build density transfer models through mathematical methods. This method increases the subject's feedback function to the environment information. Compared with the previous models, the realization of the feedback function is conducive to the realization of the dynamics of the system and the reliability of the prediction. Although the effect is good for a certain city, it ignores the spatial diffusion effect of criminal activities and has strict data requirements. [0003] The second approach is to abstractly ...

AI Technical Summary

Problems solved by technology

[0004] Although the existing crime scene modeling methods can use the extracted criminal features to fuse the interaction between the criminal subject and the environment to model the criminal network, there are two shortcomings: First, at the data level, although a bunch of heterogeneous data Throwing it into the deep learning model may be able to get a good prediction effect, but the interaction and interpretability between the features will be lost, and if you change the data set, the result will be uncertain

Second, at the method level, using the extracted features for regression analysis or predicting the crime rate in the space-time dimension, the existing models have made strong parametric assumptions, which will lead to missing crimes and different types of crimes and crimes. Dynamic interaction between locations, and relying on current static historical data, while ignoring the variability of a crime event in a dynamically changing crime scene

Images