Article Code : 13971108167191(DOI : 10.7508/jist.2018.04.004)

Article Title : DBCACF: A Multidimensional Method for Tourist Recommendation Based on Users’ Demographic, Context and Feedback

Journal Number : 24 Autumn 2018

Visited : 1656

Files : 682 KB

List of Authors

  Full Name Email Grade Degree Corresponding Author
1 Maral Kolahkaj Faculty Member PhD
2 Ali Harounabadi Assistant Professor PhD
3 Alireza Nikravan shalmani Assistant Professor PhD
4 Rahim Chinipardaz Professor PhD


By the advent of some applications in the web 2.0 such as social networks which allow the users to share media, many opportunities have been provided for the tourists to recognize and visit attractive and unfamiliar Areas-of-Interest (AOIs). However, finding the appropriate areas based on user’s preferences is very difficult due to some issues such as huge amount of tourist areas, the limitation of the visiting time, and etc. In addition, the available methods have yet failed to provide accurate tourist’s recommendations based on geo-tagged media because of some problems such as data sparsity, cold start problem, considering two users with different habits as the same (symmetric similarity), and ignoring user’s personal and context information. Therefore, in this paper, a method called “Demographic-Based Context-Aware Collaborative Filtering” (DBCACF) is proposed to investigate the mentioned problems and to develop the Collaborative Filtering (CF) method with providing personalized tourist’s recommendations without users’ explicit requests. DBCACF considers demographic and contextual information in combination with the users' historical visits to overcome the limitations of CF methods in dealing with multi- dimensional data. In addition, a new asymmetric similarity measure is proposed in order to overcome the limitations of symmetric similarity methods. The experimental results on Flickr dataset indicated that the use of demographic and contextual information and the addition of proposed asymmetric scheme to the similarity measure could significantly improve the obtained results compared to other methods which used only user-item ratings and symmetric measures.