[Zurück]

L. Glatzer, J. Neidhardt, H. Werthner:
"Automated Assignment of Hotel Descriptions to Travel Behavioural Patterns";
Vortrag: ENTER Conference 2018, Jönköping, Sweden; 24.01.2018 - 26.01.2018; in: "Information and Communication Technologies in Tourism 2018", Springer, (2018), ISBN: 978-3-319-72922-0; S. 409 - 421.

The amount of people using online platforms to book a travel accommodation has grown tremendously. Hence, tour operators implement recommender systems to offer most suitable hotels to their customers. In this paper, a method of using hotel descriptions for recommendation is introduced. Different natural language processing methods were applied to pre-process a corpus of hotel descriptions. Further, three machine learning approaches for the allocation of hotel descriptions to travel behavioural patterns were implemented: clustering, classification and a dictionary-based approach. The main results show that clustering cannot be used in this context since the algorithm mostly relies on the operator-dependent structure of the descriptions. Supervised classification achieves the highest precision for six travel patterns, whereas the dictionary approach works best for one pattern. In general, the results for the different travel patterns vary due to the unequally distributed data sets as well as various characteristics of the patterns.

The amount of people using online platforms to book a travel accommodation has grown tremendously. Hence, tour operators implement recommender systems to offer most suitable hotels to their customers. In this paper, a method of using hotel descriptions for recommendation is introduced. Different natural language processing methods were applied to pre-process a corpus of hotel descriptions. Further, three machine learning approaches for the allocation of hotel descriptions to travel behavioural patterns were implemented: clustering, classification and a dictionary-based approach. The main results show that clustering cannot be used in this context since the algorithm mostly relies on the operator-dependent structure of the descriptions. Supervised classification achieves the highest precision for six travel patterns, whereas the dictionary approach works best for one pattern. In general, the results for the different travel patterns vary due to the unequally distributed data sets as well as various characteristics of the patterns.

http://dx.doi.org/10.1007/978-3-319-72923-7_31

https://link.springer.com/chapter/10.1007/978-3-319-72923-7_31