Improving Self-Adaptation by Combining MAPE-K, Machine and Deep Learning

Abstract

Monitoring, Analyzing, Planning, and Execution share knowledge and build a favorable approach in the form of a loop (MAPE-K). However, this proposed reference model is not efficient for large self-adaptations. Moreover, the failure of the analyzer component to keep up with the current expansion of data is one of the reasons that making the MAPE-K loop consumes a lot of time and resources. We suggest a hybrid learning dataflow design for the analysis phase that combines Machine and Deep Learning techniques to enhance the accuracy of the Analyzer component in less time.