A Robust Service Recommendation Scheme

Abstract

In service computing, the quality of service (QoS) has been used to distinguish different services. Many service recommendation schemes predict how a customer might rate the QoS of various services. Based on the predicted ratings, they recommend services to the customer. Most of these schemes do not consider the unfair rating problem. As the QoS rating of a service can determine whether the service is chosen by a customer, malicious users and services might explore the weakness of the existing schemes in handling unfair ratings to gain commercial advantage. This paper proposed a service recommendation scheme that is robust against unfair rating. When predicting a customer's QoS rating for a service, the proposed scheme takes into account of the ratings given to the service by the users that are similar to the customer, the ratings that the service gained from the typical users and the own experience of the customer. Experiments with the proposed scheme show that (a) the scheme has good prediction accuracy, and (b) it can counter the manipulations by the malicious users and services effectively.