Automatic parameter tuning framework for performance diagnosis report

For ICT system operators, performance diagnosis reports that summarize the status and problems of systems in an easy-to-understand manner are important information for operating systems stably and maintaining normal conditions. In making the report, each analysis technique used for diagnosis needs its parameters tuned properly in accordance with the environment of the system. However, it is difficult for system operators to tune these parameters because they usually seem to not have enough knowledge on analysis. In this paper, we propose a framework for automatic parameter tuning that updates the parameters of analysis techniques by analyzing comments corrected by system operators in reports written in natural language. By using the framework, we confirmed that a system operator who does not have enough knowledge on analysis can easily tune the parameters of analysis techniques in accordance with the system environment.