实验技术与管理2025，Vol.42Issue(8)：62-72,11.DOI:10.16791/j.cnki.sjg.2025.08.009
犯罪组织的图数据挖掘实训设计
Experimental design for graph data mining on criminal organizations
摘要
Abstract
[Objective]The advent of the era of big data intelligence has introduced big data characteristics as activity and patterns of action of criminal organizations,which are concealed in substantial data and have increased destructive power.Therefore,this poses novel challenges to public safety management and urgently requires the mastery of techniques for analyzing the data of criminal organizations for data support in police activities such as case investigation and evidence collection.Aiming at the lack of practical teaching and training in the new research field of criminal network analysis,a training program for graph data mining of criminal networks has been designed.[Methods]First,reviewing the history of research and development in this field and summarizing its research trends demonstrated that intensive high-tech support is needed for research.In response to the lack of a unified analysis process and a standardized general framework for criminal network analysis,a data-driven interactive training and analysis framework for criminal networks has been constructed based on relevant research and police practice,achieving the integration of data analysis theory and technical tools with police work.Next,based on a discussion of the five sources and three methods of collecting data on criminal organizations,and by analyzing the basic binary structure of criminal activities,students were guided to expand the graph concerning multiple relationships and attributes viewpoints.The formal description and modeling of criminal organizations were performed using sets,which can establish a new perspective for students to understand complex criminal organizations.Meanwhile,named entity recognition and entity association extraction strategies were studied for both structured data and unstructured text to complete the construction of criminal networks.The cooccurrence strategy was used to consider the cooccurrence situation.Considering the cooccurrence analysis of report records,an experimental method was proposed to extract cooccurrence-based entity associations.This method considers crime-related locations,files,etc.,where criminal entities collectively appear,as the data basis for constructing criminal entity associations.Finally,based on the topological properties of complex networks,analysis indicators for the network characteristics of criminal organizations were constructed.The characteristics and functions of these indicators were examined,and the network characteristics of criminal organizations can be analyzed quantitatively from two perspectives:global overview features and local attribute distribution.[Results]To facilitate the observation of the unique characteristics of criminal networks,a comparative display experiment was conducted between a real Caviar criminal network data example and complex network models on the same scale to reveal some complex network characteristics and analysis processes of criminal networks.The experimental results show that the Caviar drug trafficking network has several leaf nodes,and the distribution of nodes in the middle of the network is relatively dense,which is not the case in other theoretical models.The random graph generates more connected edges with a low connection probability compared to other graphs,indicating that the probability of establishing connections between people is not easy,and criminal organizations need specific rules to establish effective structures.Moreover,from a global perspective,the Caviar drug trafficking network has the shortest network diameter,shortest average path length,and highest clustering coefficient;this indicates that the Caviar drug trafficking network has high efficiency and strong cohesion.In addition,concerning the local characteristics perspective,the node degree distribution indicates that the Caviar drug trafficking network is flatter,whereas the shortest path distribution demonstrates that it embodies the"six degrees of separation"theory well.The node betweenness distribution shows that criminal organizations concentrate communication on a small number of individuals.The local aggregation coefficient distribution indicates that there are multiple aggregation modes within it,and the edge betweenness distribution indicates that communication within the Caviar drug trafficking network can rely on a small number of edges and nodes.Furthermore,the centrality of the network degree has a notable impact on the connectivity and other characteristics of the Caviar drug trafficking network.[Conclusions]Through this practical training program,students can quickly master the basic processes,methods,and practical skills of network analysis of criminal organizations,improve their level of comprehensive analysis of criminal organizations,and enhance their ability to use network science and data mining technology effectively for complete police work,providing intellectual support for maintaining national and social public safety.关键词
犯罪网络分析/图挖掘/复杂网络/图论Key words
criminal network analysis/graph mining/complex network/graph theory分类
信息技术与安全科学引用本文复制引用
朱涛,高光亮,夏玲玲,梁广俊..犯罪组织的图数据挖掘实训设计[J].实验技术与管理,2025,42(8):62-72,11.基金项目
国家自然科学基金项目(72401110) （72401110）
2023年度江苏高校哲学社会科学研究一般项目(2023SJYB0472) （2023SJYB0472）
