A Heuristically Optimized Complex Event Processing Engine for Big Data Stream Analytics

Abstract

This paper describes a Big Data stream analytics platform developed within the DEWI project for processing upcoming events from wireless sensors installed in a truck. The platform consists of a Complex Event Processing (CEP) engine capable of triggering alarms from a predefined set of rules. In general these rules are characterized by multiple parameters, for which finding their opti- mal value usually yields a challenging task. In this paper we explain a methodol- ogy based on a meta-heuristic solver that is used as a wrapper to obtain optimal parametric rules for the CEP engine. In particular this approach optimizes CEP rules through the refinement of the parameters controlling their behavior based on an alarm detection improvement criterion. As a result the proposed scheme retrieves the rules parameterized in a detection-optimal fashion. Results for a cer- tain use case – i.e. fuel level of the vehicle – are discussed towards assessing the performance gains provided by our method.