Abnormal Condition Monitoring Based on Stacked Sparse Autoencoder

In recent years, abnormal condition monitoring has provoked vast amount o attention and research from multiple disciplines. Especially in the micro-service architecture based on the cloud platform, abnormal condition monitoring has emerged as a powerful instrument for improving the stability of service. However, along with the development of geometric growth in the number of micro-service, due to the lack of systematic methods, it is difficult to effectively analyze and extract valuable information from the large amount of data collected, the abnormal monitoring in cloud platform O&M is facing new difficulties and challenges. Traditional operation and maintenance work mainly relies on human experience to analyze a large number of indicators to determine whether there is a failure, which is very inefficient and relies heavily on expert opinions. To address these problems, in this paper, we propose a abnormal condition monitoring method based on stacked sparse autoencoder for cloud platform. Our method aims to model the operating state based on the historical experience with deep autoencoder, then the constructed model is used to analyze the current status of the service and predict the possible abnormal condition. Experimental results on real-world datasets demonstrate the effectiveness of our method compared to the traditional methods.