Providing a new similarity measure for Tourism context-aware recommender systems

Travel recommender systems are used to make POIs recommendations to target users. One of the most important algorithms used in recommender systems is the Neighborhood-Based Collaborative Filtering algorithm which finds the closest neighbors to target users or target items. The Similarity measure is a main factor to calculate similarities in Neighborhood-Based Collaborative Filtering algorithms while there are many problems in such algorithms which affect calculating similarities among users. One of the most successful similarity measures is called CJACMD which is a combination of three measures: Cosine, Jaccard, and MMD. It has been tried in this research to decrease the problems of similarity computation by advancing similarity measure CJACMD. To solve the problems of similarity measures, contextual information and demographic information have been used. Proposed similarity measure is called C.D.CJACMD . The results of experiments over the collected data set from Flickr show that the precision and recall of produced recommendations by C.D.CJACMD, in comparison with CJACMD have been improved ۱.۶ times and ۱.۲ times respectively.

Keywords:

Travel recommender system , Similarity measure , Contextual information , Demographic information