SERVICE RECOMMENDATION VIA EXPLOITING GLOBAL AND LOCAL TRUST METRICS

Recommending users with trustworthy services is a fundamental need in service-oriented environments. Various reputation-based methods have been proposed to address the issue of service trust evaluation. They typically yield a unique and global trust score for each service by aggregating the ratings on it awarded by the user community. However, this global trust score could be inconsistent with an individual’s personal opinion on the service, especially when the service is highly controversial. To attack this issue, local trust metrics have been developed, aiming to measure trust on a service more personally and accurately. However, local trust metrics, which employ trust propagation in trust networks, may fail when the judging user has no trust chains to reach the target service. By exploiting both global and local trust metrics, this paper proposes a hybrid trust measure for trustworthy service recommendation. We firstly present a global trust metric and a local trust metric different from conventional ones, and then develop a reasonable strategy for combining them to predict a user’s trust on a service. Evaluations show that our proposed method significantly outperforms the other three methods in service-oriented environments with social networks.