Multi-criteria tensor model for tourism recommender systems
Introduction
As the valuable information available on the Internet and the number of its users have increased hugely in the last decade, the amount of information provided to any query on the Web using a search engine or other application is often overwhelming. In turn, users need a lot of energy and time to find information useful for them. Intelligent systems using personalized information have been studied as a way to cope with this overload and provide an intellectually manageable number of possible recommendations (Walek & Fojtik, 2020). In the tourism industry, such recommender systems automatically extract tourists’ preferences through analysis of their explicit or implicit feedback and match the features of tourism items with their needs (Cai et al., 2018, Esmaeili et al., 2020).
Collaborative filtering (CF) is one of the most well-known and frequently used methods to recommend items in various fields. Traditional CFs are typically based on a single type of rating score. Whereas, in the case of restaurant or hotel recommendation, ratings of multiple aspects (e.g., overall, staff, service, or atmosphere) can often be collected and reflect various characteristics of the restaurant or hotel (Turk & Bilge, 2019). Indeed, in online review platforms such as Tripadvisor, Hotel.com, and Booking.com, restaurants or accommodations are often evaluated for multiple aspects, unlike movies and books. Such multiple rating data is a source of rich information to provide personalized restaurant or hotel recommendations (Fu, Liu, Ge, Yao, & Xiong, 2014). However, it is a non-trivial task to reflect the multiple ratings into recommendation services due to the unique features of multi-aspect user reviews. Moreover, the task becomes more complicated when the features have inter-relation with other factors such as spatial and temporal context (Salehan et al., 2017, Viktoratos et al., 2018, Zhang et al., 2018, Wang and Yi, 2019). In order to use multiple factors in a recommendation, multi-criteria recommender systems have been studied. However, most of the existing research (Adomavicius et al., 2011, Zheng, 2017) considered multi-criteria independently or sequentially. Whereas, we simultaneously reflect the multi-criteria ratings in rating prediction by using a “single tensor” that keeps an inherent structure of and interrelations between multi-criteria directly.
On the other hand, according to Lee (2016), cultural difference is often considered as a barrier to technology transfer. Moreover, Jung, Lee, Chung, and tom Dieck (2018) pointed out that information systems in the tourism industry are mostly affected by cultural factor. Therefore, researchers (Chen and Pu, 2008, Tang et al., 2011, Berkovsky et al., 2018) have examined cultural influences on recommender systems. Hofstede (1980) defined “culture” as “the collective programming of the mind which distinguishes the members of one human group from another.” Moreover, he and his colleague distinguished countries by five cultural dimensions: masculinity/femininity; power distance; time orientation; uncertainty avoidance; and individualism/collectivism (Hofstede & Bond, 1988). Among these five dimensions, recommender systems relate to uncertainty avoidance and collectivism (Hong, An, Akerkar, Camacho, & Jung, 2019). The uncertainty avoidance links with the purpose of recommender systems to reduce overloading information and alleviate uncertainty on decision making (Choi, Lee, Sajjad, & Lee, 2014). Collectivism associates with the recommendation functionality based on a collaborative filtering algorithm that utilizes the preferences of users similar to an active user. However, few studies have focused on the cultural influence in the item recommendation despite its substantial impact. Even existing studies (Choi et al., 2014, Chen and Pu, 2014, Chu and Huang, 2017) mainly analyze cultural differences in recommendation results through user surveys, rather than applying the cultural factors to user preference modeling and rating prediction. In contrast, the proposed model considers cultural differences directly and uses them in rating prediction.
In summary, there have been few studies that apply cultural differences to user preference modeling and evaluate its impacts on recommendation performance. To the best of our knowledge, this study is the first research work to consolidate multi-criteria ratings and a cultural factor into a single model for tourism recommendation. Moreover, the proposed model enables us to preserve the inherent structure of and interrelations between various factors (i.e., the multiple ratings and cultural factor). By using tensor factorization, the proposed model is approximated to predict user ratings for multi-criteria. Lastly, cultural differences are analyzed via the results of experiments designed to consider the cultural factor. In this regard, our primary contributions are as follows.
- •
Contrary to other related work, we consolidate user preferences along with multiple rating and cultural factor simultaneously.
- •
The proposed model outperforms other well-known techniques in terms of the recommendation prediction and its stability.
- •
This model can be easily applied to other domains such as hotel and point-of-Interest recommendations if the multiple ratings for a recommended item are collected.
- •
Like other related work, experimental results show that classifying cultures into Western and Eastern groups is an effective manner, especially to improve recommendation performance in this study.
Section snippets
Related work
Cultural difference has been conceived as one of the essential factors in the tourism research field. According to Jung et al. (2018), tourists’ cultural background relates to the experience they want. In addition, tourism supported by information technology is becoming more international along with the increasing number of travelers from various countries (Li, 2014). Likewise, tourists from different cultures have a variety of rating behaviors (Chu & Huang, 2017). Many studies have shown
Multi-criteria tensor model
This section introduces consolidation models to reflect multi-criteria and cultural factors in the context of restaurant recommendation. A tensor factorization predicting missing ratings in the models is also explained.
Dataset
This section compares our real-world dataset with the other datasets of the studies reviewed in Section 2. Table 3 shows the statistical information of all the mentioned datasets. The “Ques.” means that the dataset is gathered by a questionnaire. Half of the relevant studies (excluding our previous work) haven’t used rating scores generated by users to analyze cultural differences in recommender systems. Although two other works used rating scores, they only considered a small number of
Influences of multiple rating and cultural group factors
This section assesses how multi-criteria ratings and cultural groups affect the predictive performance of restaurant recommendation. Also, we evaluate the proposed methods based on user and country models to analyze the influences of cultural differences on the recommendation performance.
Conclusion
Recommender systems extract travelers’ preferences regarding tourism facilities such as restaurants, hotels, and museums by analyzing their explicit or implicit feedback containing their interests. Recently, famous online review platforms for tourism items often gather multi-criteria ratings from their users who come from different cultures. However, as shown in our experiments, it is not a trivial task to reflect the multi-criteria and cultural factors into recommendation services due to their
CRediT authorship contribution statement
Minsung Hong: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Writing - original draft, Writing - review & editing. Jason J. Jung: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (NRF-2017R1A2B4010774, NRF-2018K1A3A1A09078981).
References (52)
- et al.
Itinerary recommender system with semantic trajectory pattern mining from geo-tagged photos
Expert Systems with Applications
(2018) - et al.
The influence of national culture on the attitude towards mobile recommender systems
Technological Forecasting and Social Change
(2014) - et al.
A novel tourism recommender system in the context of social commerce
Expert Systems with Applications
(2020) - et al.
The confucius connection: From cultural roots to economic growth
Organizational Dynamics
(1988) - et al.
Multi-sided recommendation based on social tensor factorization
Information Sciences
(2018) - et al.
Robustness analysis of multi-criteria collaborative filtering algorithms against shilling attacks
Expert Systems with Applications
(2019) - et al.
Combining community-based knowledge with association rule mining to alleviate the cold start problem in context-aware recommender systems
Expert Systems with Applications
(2018) - et al.
A hybrid recommender system for recommending relevant movies using an expert system
Expert Systems with Applications
(2020) - et al.
New recommendation techniques for multicriteria rating systems
IEEE Intelligent Systems
(2007) - et al.
Multi-criteria recommender systems
Group recommendation: Semantics and efficiency
PVLDB
Demographics, weather and online reviews: A study of restaurant recommendations
Measuring service quality in restaurants: an application of the servqual instrument
Hospitality Research Journal
Experiments on user experiences with recommender interfaces
Behaviour & IT
Cultural difference and visual information on hotel rating prediction
World Wide Web
Balancing preferences, popularity and location in context-aware restaurant deal recommendation: A bristol, cardiff and brighton case study
A multi-criteria recommender system for tourism using fuzzy approach
Journal of Soft Computing and Decision Support Systems
Tensor methods and recommender systems. WIREs
Data Mining and Knowledge Discovery
User preference learning with multiple information fusion for restaurant recommendation
Cultural differences between east asian and north american in temporal orientation
Review of General Psychology
A neural networks approach for improving the accuracy of multi-criteria recommender systems
Applied Sciences
Culture and organizations
International Studies of Management & Organization
Cited by (35)
An investigation of the exposure effect of recommender systems in hospitality
2024, Decision Analytics JournalDSER: Deep-Sequential Embedding for single domain Recommendation
2022, Expert Systems with ApplicationsCitation Excerpt :The performance was much worse than the automatic way. Recommendation systems offer items by estimating the user’s preference (Hong & Jung, 2021). In the literature, they are classified into three base categories: content-based, collaborative filtering, and hybrid approaches.
A novel Sequence-Aware personalized recommendation system based on multidimensional information
2022, Expert Systems with ApplicationsCitation Excerpt :This pattern plays a vital role in the POI suggestions, as POIs in the grouping can influence the succeeding POI (Tang & Wang, 2018). Therefore, evaluating user behavior over time necessitates methods such as Sequential Pattern Mining (SPM) (Quadrana & Cremonesi, 2018; Hong & Jung, 2021). Due to the importance and complexity of recommender systems, there is a need to maximize their accuracy, which necessitates using hybrid methods, especially in the concurrent surge of social networking (Chaudhari and Thakkar, 2020; Dara et al., 2020).
Blockchain-based recommender systems: Applications, challenges and future opportunities
2022, Computer Science ReviewCitation Excerpt :Recommender systems (RSs), initially introduced to address the problem of improving the customer experience and retention in e-commerce sites [1] has since become a ubiquitous and often anticipated functionality of many online interactions, from movie and song recommendations [2] to applications related to tourism [3], social networks [4], health [5], energy and smart cities [6], and many more.
Leveraging item attribute popularity for group recommendation
2024, International Journal of System Assurance Engineering and Management