ASW-RTS: An Online Anomaly Detection Method for Software Performance
Xie Weiq · Aerospace Control · 2015
In the software health management system,the performance of related component is required to be online detected for determining the health status of the component. Regarding the performance degradation of tested component,a ASW- RTS algorithm with self- adaptive strategy is designed,which is based on the AR( p) model and combined with the accuracy of time series rolling algorithm. This algorithm can be used to on- line detect the innovation outlier of the component performance data with high accuracy,and then an on- line detection process based on the ASW- RTS algorithm is designed. The experiment results show that the detection accuracy using this algorithm is improved by 50%,and the performance degradation of tested component can be online detected. Due to the experiment performance,the correctness and effectiveness of using ASW- RTS algorithm are verified.