Complex Event Detection for Long Process

The traditional CED (Complex Event Detection) methods were achieved in memory by creating a data structure. But the memory will become a choke-point for the huge data structure when orienting long process and numerous data. So large amounts of data must be stored in external memory, and transferred to memory when needed. In this paper, we present TSH (Hash by object ID based on time slice model) storage strategy and instance map structure to ensure small I/O cost when transferring instances needed to memory. In addition, we also propose an incremental method to accelerate the matching process. Finally, the validity of the methods presented in this paper is proved by experiments on simulated data sets.