Multi-indicators prediction in microservice using Granger causality test and Attention LSTM

Suozhao Ji, Wenjun Wu, Yanjun Pu

Abstract: In the field of microservice, accurate indicator prediction is very important, which is helpful for service monitoring and anomaly detection. In many cases, it is difficult to accurately predict by the indicator itself, and other related indicators need to be imported to help predict. In traditional multi-indicator predicting, the related indicators are known or the amount is small, which is relatively easy to obtain. But there are many service indicators and the relationship between the indicators is constantly changing, so new methods need to be used to quickly and accurately find the related indicators in the mass of indicators. We combine Granger causality test and Attention LSTM time series prediction model to quickly find related indicators in microservice scenarios and participate in prediction. The experimental results show that our method can effectively improve the accuracy of indicator prediction.