Trust and Reputation Modelling for Tourism Recommendations Supported by Crowdsourcing

Abstract

Abstract. Tourism crowdsourcing platforms have a profound influence on the tourist behaviour particularly in terms of travel planning. Not only they hold the opinions shared by other tourists concerning tourism resources, but, with the help of recommendation engines, are the pillar of personalised resource recommendation. However, since prospective tourists are unaware of the trustworthiness or reputation of crowd publishers, they are in fact taking a leap of faith when then rely on the crowd wisdom. In this paper, we argue that modelling publisher Trust & Reputation improves the quality of the tourism recommendations supported by crowdsourced information. Therefore, we present a tourism recommendation system which integrates: (i) user profiling using the multi-criteria ratings; (ii) k -Nearest Neighbours (k -NN) prediction of the user ratings; (iii) Trust & Reputation modelling; and (iv ) incremental model update, i.e., providing near real-time recommendations. In terms of contributions, this paper provides two different Trust & Reputation approaches: (i) general reputation employing the pairwise trust values using all users; and (ii) neighbour-based reputation employing the pairwise trust values of the common neighbours. The proposed method was experimented using crowdsourced datasets from Expedia and TripAdvisor platforms.

Analysis and prediction of hotel ratings from crowdsourced data

“…Therefore, reputation is a collective measure of trustworthiness based on the referrals or ratings provided by community members (Jøsang et al, ). Trust and reputation are distinct, but intrinsically linked concepts, for example, “I trust that restaurant because it has good reputation.” The trust and reputation guest modeling has been implemented in hotel recommendation engines using fuzzy membership functions (Jøsang, Guo, Pini, Santini, & Xu, ), or the trustworthiness between common neighbors (Leal, Malheiro, & Burguillo, ).…”

“…The majority of the surveyed rating‐based profiling approaches use entity‐based modeling. In addition, while Wu et al () and Leal, Malheiro, et al () model the trust, the Jøsang et al () and Leal et al () explore the reputation in hotel recommendation supported by crowdsourced ratings.…”

“…Specifically, in the case of hotel recommendation, the search space can be tailored using: (a) prefiltering; (b) postfiltering; and (c) pre‐ and postfiltering. Typically, prefilters tend to rely on user context data, for example, travel type (business or leisure) and user home location (Zheng et al, ), whereas postfilters on item data, for example, the hotel value for money (price, official hotel star rating, and overall rating) and hotel sentiment value (applying sentiment analysis to the textual reviews) (Barbosa, Sánchez‐Alonso, & Sicilia‐Urban, ; Leal, ). Concerning the latter case, Barbosa et al conclude, based on the evaluation the sentiment analysis technologies applied to crowdsourced hotel reviews, that sentiment analysis postfiltering improves decision‐making both for companies and individuals.…”

“…All surveyed hotel recommendation engines use collaborative filtering for rating prediction. Specifically, Zheng et al (); Jannach et al (); Jøsang et al (); Colace et al (); Hu et al (); Farokhi et al (); Leal, González‐Vélez, et al, ; Leal, Malheiro, & Burguillo, ; Leal, Malheiro, et al, ; and Leal et al, ) use a single filtering method; Wu et al (); Nilashi, bin Ibrahim, et al, ; Nilashi, Jannach, et al, ); and Song et al () employ a combination of collaborative filtering methods, that is, mixed; and Nilashi et al () apply an ensemble of collaborative methods. In terms of processing, the majority of the surveyed hotel engines implement off‐line processing.…”

“…In terms of processing, the majority of the surveyed hotel engines implement off‐line processing. However, Nilashi, Jannach, et al (); Nilashi et al (); and Leal et al () perform on‐line processing, that is, using incremental updating. While the prerecommendation filtering has been used to reduce the search space (Colace et al, ; Zheng et al, ), the postfiltering has been explored to improve the recommendation accuracy mainly in on‐line recommendations (Leal, ).…”