Web Services Quality Prediction Based on Multivariate Time Series Analysis
Pan He, Yue Yuan, Gang Liu · 2018
Web service quality prediction helps to identify quality degradation in online system maintenance. While historical web service usage data is used to predict the service quality in the near future, the similarities in the service usage data from multiple users invoking the same service is ignored. To improve the service quality prediction accuracy, a web service quality prediction method is built considering multiple-user invocation process. A multivariate time series using vector ARMA models is used to characterize the multiple invocation process. After analyzing the similarity in the historical web service quality data, a quality prediction method is proposed based on the multivariate time series to predict the quality data for the next few time series points. Experimental evaluations were conducted to illustrate the multivariate time series model construction process and to compare the multivariate method with the univariate methods. Comparison results showed that in most test cases, the multivariate prediction method outperformed the univariate models in both MAE and RMSE.