Research On Key Technology And Applications Of Social Network Auxiliary Investigation

With the development of mobile Internet,people have been inseparable from social networks in their work and life,and accumulated a lot of social relationship data and text data.The public security departments rely more and more on these data in case investigation: using relational data,we can identify key suspects and an alyze the evolution mechanism of organizational structure;Text data can be used to detect case related topics.Therefore,it has important application value and practical significance to help investigators broaden their thinking and choose the breakthrough point of investigation.The main work of this paper includes: the research of key suspect mining,dynamic organization mining,suspect topic forensics and the design and implementation of social network auxiliary investigation system.The details are as follows:Firstly,aiming at the problem of key suspect mining,a key node mining algorithm based on multiple query nodes is proposed.In this method,the known suspects are regarded as query nodes,and the local topology is extracted.By calculating the critical degree of non query nodes in the local topology,the nodes with higher critical degree are selected for recommendation.The experimental results show that the accuracy of the key nodes identified by this method is higher than that of the traditional key node mining algorithm.Secondly,aiming at the problem of dynamic organization mining,this paper proposes a single time slice organization discovery algorithm Enhanced-Label Rank and an organization dynamic evolution tracking algorithm.Based on the Label Rank algorithm,the former introduces the complementary information of multiple adjacent time slices to enhance the effect of organization discovery on a single time slice;The latter uses the results of Label Rank enhance algorithm to identify seven organizational evolution states and generate organizational evolution chain.Experimental results show that the Enhanced-Label Rank algorithm performs better than the traditional single time slice organization discovery algorithm on multiple data sets.Thirdly,aiming at the characteristics of short sentences,few single sentence feature words and nonstandard expression in the suspect chat text,a short text topic word detection method is designed.This method realizes text topic detection through three key steps: sentence representation vector generation,short text clustering and topic word extraction.The experimental results show that the accuracy and recall rate of this method are improved by 42% and 30% respectively.Finally,based on the above research content,a social network auxiliary investigation system is designed and implemented,which provides effective technical means for key suspect mining,dynamic organization mining and suspect topic forensics.The system test results show that the system can achieve the expected goal.