A Tourist Recommendation System: A Study Case in Mexico

The present work deals with implementing tourist recommendation systems designed to predict the user preferences about a place or tourist activity in Mexico. Three recommendation systems have been proposed: two based on collaborative filtering (user and items) and the other based on demographic issues. To this aim, a corpus has been built by collecting 2,263 ratings from TripAdvisor.com about eighteen tourist places in Mexico. Experimental results show that the demographic-based recommendation system outperforms those based on collaborative filtering, obtaining a mean absolute error of 0.67 and a mean square error of 1.2980. These results also show significant improvement over a majority class baseline based on a sizeable unbalanced corpus.