An interval-based multi-objective artificial bee colony algorithm for solving the web service composition under uncertain QoS

Abstract

Most of the existing works addressing the QoS-aware service composition problem (QoSSCP) are based on the assumption of fixed quality of service (QoS) characteristics of elementary web services. However, in the real world, some QoS criteria may be imprecise for many unexpected factors and conditions. Therefore, when dealing with QoSSCP, we must consider the uncertain proprieties of QoS. Moreover, very few studies propose multi-objective solutions for solving the QoSSCP, and there is no multi-objective algorithm solving the QoSSCP under uncertain QoS, in which the non-deterministic values of the QoS attributes are expressed as interval numbers. To resolve this issue, we formulate an interval-constrained multi-objective optimization model to the QoSSCP, and we propose a novel interval-based multi-objective artificial bee colony algorithm (IM_ABC) to solve the suggested model. To deal with the interval-valued of objective functions, we define an uncertain constrained dominance relation for ordering solutions in which the performance and stability are simultaneously considered. As inspired by Deb’s feasibility handling constraints, a new interval-based feasibility technique is proposed to deal with interval constraints. In order to control the diversity of the non-dominated solutions obtained by IM_ABC, the original crowding distance of NSGA-II is extended and adopted to the uncertain QoSSCP by incorporating to it a new interval distance definition. Based on real-world and random datasets, the effectiveness of the proposed IM_ABC has been verified through multiple experiments, where the comparison results demonstrates the superiority of IM_ABC compared to the recently proposed interval-based multi-objective optimization algorithms IPMOPSO, IPMOEOA, and MIIGA as well as a recently introduced interval-based fuzzy ranking single-objective GAP approach.