Hybrid Learning System for Adaptive Complex Event Processing

摘要

In today's security systems, the use of complex rule bases for information aggregation is more and more frequent. This does not however eliminate the possibility of wrong detections that could occur when the rule base is incomplete or inadequate. In this paper, a machine learning method is proposed to adapt complex rule bases to environmental changes and to enable them to correct design errors. In our study, complex rules have several levels of structural complexity, that leads us to propose an approach to adapt the rule base by means of an Association Rule mining algorithm coupled with Inductive logic programming for rule induction.