Incremental Hotel Recommendation with Inter-guest Trust and Similarity Post-filtering

Abstract

Crowdsourcing has become an essential source of information for tourists and tourism industry. Every day, large volumes of data are exchanged among stakeholders in the form of searches, posts, shares, reviews or ratings. Specifically, this paper explores inter-guest trust and similarity post-filtering, using crowdsourced ratings collected from the Expedia and TripAdvisor platforms, to improve hotel recommendations generated by incremental collaborative filtering. First, the profiles of hotels and guests are created using multi-criteria ratings and inter-guest trust and similarity. Next, incremental model-based collaborative filtering is adopted to predict unknown hotel ratings based on the multi-criteria ratings and, finally, post-recommendation filtering sorts the generated predictions based on the inter-guest trust and similarity. The proposed method was tested both off-line (post-processing) and on-line (real time processing) for performance comparison. The results highlight: (i) the increase of the quality of recommendations with the inter-guest trust and similarity; and (ii) the decrease of the predictive errors with the on-line incremental collaborative filtering. Thus, this work contributes with a novel method, integrating incremental collaborative filtering and inter-guest trust and similarity post-filtering, for on-line hotel recommendation based on multi-criteria crowdsourced rating streams.