Research On Complex Event Rule Learning Based On Logic

In the context of data stream,it is particularly important to handle events of interest to users.The Complex Event Processing System(CEP)not only mining events of interest to the user,but also responds to the excavated events in time.The system analyzes the connection between events,utilizes the causality and aggregation of events,and finally realizes the transition from simple events to advanced events.It is very important for the definition of rules in a complex event processing system.The system matches complex events through the rule base to realize the conversion from simple events to advanced events.At present,the relatively complete rule learning system is usually built on the existing complex event processing system in home and abroad,and the rule engine is used to match complex events.The existing complex event rule learning system divi des the rules into several modules and learns each module separately.However,it cannot process events with rich semantics,and cannot use background knowledge to deal with the uncertainty.Logic-based event structure representation systems are of great i nterest because of their declarative semantics,and logic rules can be well combined with background knowledge and support event reasoning.In this case,this paper proposes a logic-based complex event rule learning method that uses logic programming to bu ild rules from simple events automatically.The main work of this paper is as follows :1)This paper studies the logic-based event rule learning method.Firstly,the deficiencies of the existing logic-based complex event rule learning methods are analyzed.Combined with the background of streaming big data processing,a logic-based online learning method for complex event rules is proposed.Based on the inductive logic programming,this paper draws on the VFDT decision tree model and uses Hoeffding bound to traverse the data stream,learning complex event rules from simple events.The design experiment is evaluated online on the real data stream and compared with other methods to prove the effectiveness of the method.2)Based on the online learning framework,combining the probabilistic statistical method with the online framework,an online learning method for complex event rules based on weight optimization is proposed based on the structural learning strategy of Markov logic network.After learning the structure of the rules,the rules are updated by weight optimization,which further improves precision.Experiments show that this method has better performance than the online learner.