M.A. Thesis:

Improving web service recommendations using trust based on user clustering

The unprecedented growth of Internet technology in recent years has created a lot of e-commerce applications; one of the most important applications is recommended systems. In this research, trust clustering has been used to increase the accuracy of the proposed items. In the proposed model, using a trust-based clustering, a service recommendation will be made. For this purpose, clustering is performed on two levels of items (the same services) and users. At the item level, clustering is based on their characteristics. But at the user level, clustering, in addition to user characteristics, also depends on the level of trust between each user pair. Trust between users is based on two explicit and implicit trust approaches. Finally, for the top N item to be offered to a user (active user), the views of other users of the cluster he belongs to, as well as the cluster items with the active user cluster. The proposed method was evaluated using RMSE criteria, coverage and F-measure. The results showed that the proposed method based on trust-based clustering for all users with a RMSE of 0.9821, a coverage of 98.97%, and an F-measure of 0.8734. Also, for the cold start users, the RMSE suggested the rate of 1.02 was equal to 81.01% and the F-measure was 0.7734, and it got the best results compared to the comparative methods

Keywords: clustering, recommender system, trust