Research On Complex Event Detection For Long Process

Large quantities of event data are generated with the EDGE (Electronic Data Gathering Equipment) such as RFID (Radio Frequency Identification) and sensor. CED (Complex Event Detection) technology can get the information users need from the numerous data, so people paid more attention to it and widely studied it these years. In order to detect the complex events which occur in a relatively short period, those traditional methods are achieved in memory by creating a data structure, adopting some optimal strategies to decrease the middle results. However, lots of complex events may occur in a very long period in our real life besides those occurring in relatively short period. In other words, there are many event streams which satisfy the characteristic of long process around us. Because the memory is limited, traditional complex event detection technologies can’t process this kind of event stream. Related research on complex event detection for long process does not exist currently. So it is urgent to research the method of detecting complex event for long process.In order to realize complex event detection for long process, based on traditional methods and combining the characteristics of the long process event stream, this thesis creates a series of data structures, and proposes the methods which are used to support detecting complex event for long process based on these structures. The emphasis of this thesis is efficient dispatch scheduling of event instances between memory and external memory. Moreover, it also proposes an algorithm to accelerate the process of complex event detection. The main contributions of this thesis mainly reside in:First, in order to store event instances arriving with event stream, this thesis proposes TSH (Hash by object ID based on timeslice model) instance storage strategy. It can provide regular event instances for complex event detection orienting long process.Second, based on TSH strategy, a structure called instance map is created to help query needed instances. It can be seen as index for the storage location of previous event instances, so this thesis realizes the algorithm of efficient dispatch scheduling of candidate instances by combining instance map with TSH strategy Third, this thesis presents an incremental matching method for matching process. With the help of timestamp prefix array, this method can directly output the result sequences by sharing subsequences, thus achieve the destination of speeding up the matching process, and then accelerate the whole process of complex event detection for long process.Finally, the methods of complex event detection for long process at the situation of time window sliding are realized by expanding the above methods.The theoretical analysis and experimental evaluations show that the methods of complex event detection for long process proposed in this thesis have great efficiency and feasibility.