Intelligent Automation, Inc. — Department of Defense SBIR Phase I: Self-adaptive monitoring systems are highly desirable and demanded in military and civilia

Description

Self-adaptive monitoring systems are highly desirable and demanded in military and civilian application scenarios. In this effort, we leverage our in-house machine learning framework, ABMiner and ontological knowledge representation workbench to address the challenging problem, and propose the self-adaptive monitoring architecture, called SAM. In SAM, we mature data mining techniques to set up a multi-feature model for the monitored application. The multi-feature monitoring model can be used to monitor the execution of the application online. ABMiner tool allows us to use various machine learning techniques, including ensembles, to select the most significant semantic features to characterize the monitored application. Moreover, we will enhance ABMiner by integrating the weighted ensemble methods to automatically adjust to drifting features inherited in software applications. Additionally, we leverage our work on ontological knowledge representation techniques to represent the execution of an application program in an understandable way to enable human experts to adjust the automatic reasoning system. Finally, we will integrate the proposed techniques in a workable self-adaptive monitoring prototype capable of checking the execution of dynamically evolving applications.