Ontology Matched Cross Domain Personalized Recommendation of Tourist Attractions

Abstract

In this era of data deluge, recommender system lists the most likely preferred items to the users. With the vast amount of information, personalization of recommendation is a challenge. Domain knowledge plays a vital role in filtering the data for personalized recommendation. Certain domains does not have sufficient history of data to provide effective recommendation to the users. In such cases, knowledge from a relative domain is transferred to make effective recommendations. The proposed cross domain recommender system deduces relatedness between domains for knowledge transfer. Grouping the users into clusters of similar tastes works best in providing recommendation in real time environment. The proposed novel clustering based transfer learning algorithm incorporates content and collaborative properties of items and users for providing cross domain recommendation. The experiments are conducted with real world dataset which show that transfer learning technique improves the efficiency of recommendation in a sparse domain.