An Experiment Study of Service Discovery Using the Extreme Learning Machine Based Approach

Abstract

Recent years have witnessed the rapid development of Web services on the internet, providing the increasing number of online services with diverse types available today. The demand of finding the target web service that meets user’s requirement is thus no longer an easy task and needs to be paid more attention. To solve this problem, this study proposed a new service discovery method, combining the Extreme Learning Machine (ELM) and Differential Evolution Algorithm (DE) to retrieve the target service. We first calculated the similarities of each service in the training set by using four different similarity measurements and obtained the corresponding DE fitness values to construct the sample vectors. Second, these vectors were used to constitute the ELM model to learn the interrelationship between the similarity scores and the DE fitness values. Finally, we simulated the discovery process on a test set. For each new query, the target service can be received through the DE fitness values that are predicted from the constructed ELM model. The experiments were conducted on a public service set and the results showed the significant implications.