Topic based context-aware travel recommendation method exploiting geotagged photos

Abstract

The popularity of camera phones and photo sharing websites, e.g. Flickr and Panoramio, has led to huge volumes of community-contributed geotagged photos (CCGPs) available on the Internet, which could be regarded as digital footprints of photo takers. In this paper, we propose a method to recommend travel locations in a city for a user, based on topic distribution of his travel histories in other cities and the given context (i.e., season and weather). A topic model is used to mine the interest distribution of users, which is then exploited to build the user–user similarity model and make travel recommendations. The season and weather context information is considered during the mining and the recommendation processes. Our method is evaluated on a Flickr dataset, which contains photos taken in 11 cities of China. Experimental results show the effectiveness of the proposed method in terms of the precision of travel behavior prediction.