Intelligent Multi-Criteria Recommendation System for Online Hotel Booking

Liu Yuting

Abstract

With the rapid growth and development of e-commerce industry, people are spending more time and more money shopping online (Corless, 2019). They visit online websites to gain information, make comparisons and make booking or purchasing decisions. They also leave reviews on the websites in the form of comments or ratings after or during their booking or purchasing, which can be referred by other visitors to the websites later (Statista, 2018). In return, they read other people’s reviews to facilitate their decision-making process, which forms an organic cycle. In order to create better user experience, many online platforms start to work on better recommendation systems which can facilitate the users in finding what they want easier and faster. Recommender System is an efficient tool for today’s e-commerce. By taking the individual’s opinion in to consideration, system can identify their content more appropriately and selectively (Polatidis, 2013).

However, there are many issues with current recommendation systems, especially for the online hotel booking platforms. Due to the overwhelming online information and the accelerating growth rate of searching platform number, the searching and booking process becomes more confusing and time-consuming (Huang, 2013). Reviews crawled from the travellers visiting sites are a common and valuable source of information for recommendation of a hotel, yet little attention has been paid as, how to present the reviews of a reviewers in an understandable format (Gavilan, 2017). Besides, the recommendation system is always based on the simple filtering logic and the ranking engine is always according to the hotels’ advertisement payment, or simply on an ascending price logic. This can cause trouble to the users, as they cannot make the best decision in the shortest time and need to take the extra effort to compare between different hotel booking websites or need to go to second, third or even further pages (Weissenberg, 2018). There is no customized design, and the system is not able to learn users’ preference as the users accumulate more and more footprints on the website.

In this paper, an Intelligent multi-criteria recommendation system is brought up to bridge the gap, which can solve the issue of the messy comments, as well as the un-customized recommendation and ranking system. Keywords extraction and sentiment analysis are conducted, by customizing the natural language processing models. Features of the users and listing hotels will be vectorised based on the keywords. By clustering the hotels and users, relationship between preferred clusters is studied by different machine learning prediction models, which is used to develop the preference learning model. For the user, when he or she goes on the website, the recommendation system will first gather all the available data of the user, find his user group and predict his preferred hotel cluster, and then rank the hotels in the cluster according to the user’s unique utility expectations.