A Novel Web Service Quality Prediction Framework Based on F-ELM

Abstract

With the prevalence of service computing and cloud computing, more and more services are emerging and running on highly dynamic and changing environments (The Web). Under these uncontrollable circumstances, these services will generate huge volumes of data, such as trace log, QoS (Quality of Service) information and WSDL files. It is impractical to monitor the changes in QoS parameters for each and every service in order to timely trigger precaution, due to high computational costs associated with the process. In order to overcome the above problem, this paper proposes a web service quality prediction method based on improved Extreme Learning Machine with feature optimization. First, we extract web service trace logs and QoS information from the service log and convert them into feature vectors. Furthermore, in order to actively respond ELM more quickly, we mine early feature subsets in advance by developing a feature mining algorithm, named FS, and apply such feature subsets to ELM for training (F-ELM). Experimental results prove that F-ELM (trained by the selected feature subsets) can efficiently lift the reliability of service quality and improve the earliness of prediction.