A situation refinement model for complex event processing

Abstract

Complex Event Processing (CEP) systems aim at processing large flows of events to discover situations of interest (SOI). CEP uses predefined pattern templates to detect occurrences of complex events in an event stream. CEP systems rely on domain experts to define complex patterns rules to recognize SOI. The task of identifying complex patterns faces with several challenges, such as the complexity of writing the pattern rules, and the need to acquire and process background information considering event stream’s real-time constraints. Developing an efficient rule mining algorithm to fine-tune the CEP pattern to recognize SOI requires the tackling of three main obstacles. First, the CEP pattern rules must be inferred by utilizing the user’s preferred context and the history of the event stream. Second, to avoid the issue of pattern complexity, the minimum number of rules must be used in the refinement process. Finally, to respond to emerging situations, the refinement task must be fulfilled in near real-time. In this work, we present a rule mining model to refine the CEP pattern rules by considering these obstacles while providing the ability to adjust the level of refinement to fit the applied scenario. This paper aims to: (1) Review the challenges associated with incorporating domain knowledge in CEP systems to improve awareness of real-world situations; (2) Present a Situation Refinement model to extract the minimal set of rules from external knowledge required to identify Situations Of Interest; (3) Demonstrate the summary update process of the event stream; and (4) Evaluate the derived rules with respect to their coverage and complexity.