An overall performance and rule-judge based service selection algorithm

Abstract

Service selection problem based on Quality of Service (QoS) aware is a multi-attribute decision making problem. Generally speaking, services which have the largest value of overall performance will be selected. However, if the top-two services have the same or approximate overall performance, but values of the QoS indicators of the service with higher overall performance are seriously unbalanced (i.e., variance of the values of the QoS indicators are larger), thus choosing the service with larger overall performance means choosing the service with more serious defects in some indicators. Therefore, the degree of balance of evaluation indicators must be considered into the final service selection. An overall performance and rule-judge based service selection algorithm (ORSSA) is proposed. Relative variance is proposed here to measure the degree of balance for candidate service/schema indicators. A simulation experiment is given, which proves the ORSSA algorithm is effective and sensitive to the variation of weights. In the final service selection, users' requirement is fully taken into account by the degree of relative balance for QoS indicators, which demonstrates the superiority of ORSSA algorithm.