Quality of Service timeseries forecasting for Web Services: A machine learning, Genetic Programming-based approach

Today, many software systems and applications are consisted of various services on the Web (Cloud). When selecting services or performing a service operation, a critical criterion is Quality of Service (QoS). Because the actual value of some dynamic QoS attributes could vary with time, there must be an approach that can accurately forecast future QoS value. In this paper, we propose to use a machine learning technique, i.e., Genetic Programming (GP), for the problem. When performing QoS forecasting, the proposed approach employs GP to evolve out a predictor, and then uses it to obtain future QoS forecasts. To test and understand the forecasting performance (accuracy) of the proposed approach, in our experiments with a real-world QoS dataset, we compare our approach with other existing QoS forecasting methods, and then prove and discuss its outperformance.