University of Twente Student Theses

Login

Hotel-related attributes and hotel perceived value: A case study in New York City based on geodata science and machine learning

Xu, Yung (2020) Hotel-related attributes and hotel perceived value: A case study in New York City based on geodata science and machine learning.

[img] PDF
2MB
Abstract:Tourism is one of the industries that drive the world’s economic growth, whereas the hotel industry has always occupied a prominent position in tourism-related industries. Visitors usually post reviews to express their impressions on hotels regarding factors both from inside and outside of the hotel. Therefore, not only hotel attributes but also built-environment attributes and reviewer-related attributes should be taken into consideration in hotel research. Reviewers score the hotel in five dimensions in the TripAdvisor platform: overall rating, location, cleanliness, service, and value. Except for value, all the other indicators can be judged from specific aspects. While value, here to be more specific, hotel perceived value, is a more compound concept, whose score could be influenced by multiple factors. Hotel perceived value is a kind of intangible asset of the hotel. It is less studied and often overlooked but plays an important role in the hotel industry. In this study, the relationship between the hotel-related attributes and hotel perceived value is inspected by taking New York City as a case study area. We adopted TripAdvisor platform and NYC Open Data as the data sources, and applied geodata science in data processing, i.e., collecting hotel information data via web crawling, and transforming reviewer address data into coordinates via geocoding. Machine learning was involved in predicting hotel perceived value. Nine machine learning methods are compared: Ridge classifier, Logistic regression, Decision Tree classifier, Bagged Decision Tree, Random Forest classifier, Gradient Boosting Machine, XGB classifier, Support Vector classification, and K-Nearest Neighbors. Among them, the XGB classifier performed best. Indicators' accuracy, F1 score, recall score, and precision score are as high as 0.8. The XGB classifier feature importance function picks hotel ranking and negative review amount as the most prominent features regarding hotel perceived value. The built environment occupies the largest proportion of hotel-related attributes, more than half. However, these issues are suggested as not very significant in this study; only the accessibility/convenience to restaurants, attractions, and airports slightly show the position. Attributes related to the hotel itself display their significance in the importance ranking. One of the reviewer-related attributes, the number of cities that reviewers come from, is relatively important rather than the rest two. The reliability of the hotel's perceived value on the TripAdvisor site is also worthy of consideration. Suggestions for hotel managers are provided that they are supposed to improve the hotel ranking and remove the influence of the negative comments by responding more to these negative feedbacks in order to enhance hotel perceived value. Nevertheless, cautions are needed when generalizing the results from this study, given the potential presence of multicollinearity, which, however, does not affect the overall performance of the prediction. In future work, issues related to reviewer classification, review text analysis, reviewer address, and hotel classification could be considered to build up a more complete and convincing hotel perceived value research framework.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/85215
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page