GAN-ASD: Precise Software Aging State Detection for Android System Based on BEGAN Model and State Clustering

摘要: Software applications may become no response or stop running due to performance degradation, system crashes, program cumulative failures, after long-term execution in Android system. These phenomena have been validated be common mobile systems which is caused by software aging. To handle the aging dilemma, rejuvenation an efficient way. In order make more efficient, identifying state of precisely key point. this paper, we propose a novel detection method based on Boundary Equilibrium Generative Adversarial Network (BEGAN) and clustering technology, named as GAN-ASD. The has three phases: Firstly, Interpolation Clipping Processing used processes time series dataset constituted sample Aging Indicators. Secondly, according dataset, BEGAN generation will fit user’s usage habits generate characteristics. At last, use generative train K-Means model. With trained model, can determine whether current enters into remains normal state. validate effectiveness GAN-ASD, two evaluation criterion our comparison experiment. One coefficient (RC) evaluates user experience other one frequency (RF) cost. results show that performs better than fixed-interval random operations.