OCM - An Optimized Model to Eliminate Invalid Data in Complex Event Processing

Abstract

Complex Event Processing (CEP) is a branch of stream data mining, which is usually used to discover event patterns. Generally, a complex event processing engine uses nonprocedural declaration language and state machine to define event patterns. It generates synopses of universe data by sliding window model, therefore could identify event patterns which users care about from rapidly changing and potentially infinite event stream in a relatively short time. Among all the data streams, monitoring data flow is easy to describe by events, so CEP is suitable for supervision or decision control for business system. However, CEP does not support backtracking to eliminate invalid data, which limits CEP in supervision or decision control applications. The traditional complex event processing try to solve this problem by defining more complicated event patterns to avoid invalid results, the side effect of which would be heavier workload onto the engine. For this problem, we propose an optimized complex-event-processing model OCM. OCM will decompose complicated event patterns into small ones according to their relations, and distribute these patterns to more than one CEP engines to hierarchically filter invalid data. It can eliminate invalid data with better performance than the traditional approaches. We have also implemented the prototype and designed experiments to test the effectiveness of OCM. We have applied this model into Aviation Application in EU FP6 project Bridge.