Complex Event Detection In Wireless Sensor Networks Based On Graph Approximate Matching

Complex event detection is a main research emphasis of complex data management. Recent years, with the development of correlative technologies, user behavior detection and recognition technology based on wireless sensor networks is attracted much attention of many researchers and scholars.Currently the data model select and search algorithm in the relevant research is poor performance. This thesis proposes a new method of user behavior detection based WSN environment. The method divides data model into two layers:event graph and behavior graph model and a new graph similar algorithm is proposed. Details are as follows:First, from the mian design requirements, equipment configuration and other environmental needs, given the user behavior recognition system design schema based on WSN, and deployment this system in the laboratory.Second, based on the data snapshot graph model, modeling the event using the data incremental graph algorithm.Because the traditional method take insufficient with the different importance of the feature structural. This part applied a novel method rank the feature structure. And then we use LCS and SED to calculate the similarity between sequences. According to the different importance in different features, this thesis peoposed an improved weighted LCS and SED algorithm.Third, this thesis use of multiple sequence alignment algorithms modeling behavior graph based on the event graph sequences. In the behavior graph query phase, this thesis proposed a novel feature selection algorithm based on similar structural. After that,we use the same method calculate the similarity between behavior graph, and get the different user behavioral patterns.The experimental results shows that the modeling and query algorithms has high efficiency compared with tradition methods.