Applying Genetic Programming for Time-Aware Dynamic QoS Prediction

Abstract

A common defect of most current QoS information exposed is that they are static and did not consider some facts (e.g., Different calling time points) that can cause the actual values of some types of QoS to vary. A solution for such issue is to develop a valid forecasting mechanism able to predict future dynamic QoS values. In the past, several such forecasting approaches already have been developed. However, many of them are based on fixed statistical models and the others' prediction generation process is not understandable and observable. In this paper, we propose to employ Genetic Programming (GP), which is a powerful predictor searching/learning paradigm with very great performance reports in many other forecasting applications and never being applied to dynamic QoS forecasting yet. In this work, we study applying GP to the defined time-aware QoS forecasting problem and we report our experiment results showing and verifying the applicability and performance of GP to the problem.