Title: Modelling and exploring historical records to facilitate service composition

Authors: Jian Wu; Liang Chen; Yanan Xie; Lichuan Ji; Zhaohui Wu

Addresses: College of Computer Science, Zhejiang University, Hangzhou, China ' College of Computer Science, Zhejiang University, Hangzhou, China ' College of Computer Science, Zhejiang University, Hangzhou, China ' College of Computer Science, Zhejiang University, Hangzhou, China ' College of Computer Science, Zhejiang University, Hangzhou, China

Abstract: Along with a continuously growing number of Web services, how to locate appropriate Web services to complete the task of service composition is becoming more critical. Differing from most recent studies which mainly focus on functional and non-functional properties, we mine nuggets from the Historical Service-composition Dataset (HSD), which carries related users' past experiences. In this paper, a graph mining based recommendation approach is presented to facilitate the process of service composition. In particular, we first extend the graph mining approach gSpan to recognise Frequently Used Web Services (FUWS) with their connecting structures from HSD. Then, according to the records in HSD, which share the same FUWSs with user's partially composed service, a bunch of Web services with higher probability is recommended automatically as candidates. Finally, the skyline approach is adopted for optimal composite service selection with consideration of overall quality of services (QoS). Furthermore, experiments based on 1,530 real Web services demonstrate the effectiveness and efficiency of our approach.

Keywords: web services; recommendation systems; graph mining; QoS; quality of service; skyline; modelling; historical records; service composition; composite service selection.

DOI: 10.1504/IJWGS.2014.058760

International Journal of Web and Grid Services, 2014 Vol.10 No.1, pp.54 - 79

Received: 08 Jan 2013
Accepted: 08 Apr 2013

Published online: 29 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article