Hotel recommendation system by bipartite networks and link prediction

Abstract

With the rapid growth of the Internet in recent years, online social media has become very important for people. People often use social media tools to communicate and share their ideas or experiences regardless of time and place. One of the areas where the use of these tools is widespread is tourism. It is one of the hardest tasks to find a suitable hotel for travellers. There are many websites where accommodation services of the tourism enterprises are evaluated. People share their experiences about the hotels they stayed through these websites. Positive or negative comments are an effective factor in hotel selection. The purpose of this study is to construct a hotel recommendation system based on user’s location using online hotel reviews. The reviews crawled from TripAdvisor.com were filtered according to their total scores, and a dataset was obtained consisting of users and reviews of their liked hotels. For the recommendation task, first, three different bipartite networks consisting of users and hotels were modelled, which were global, country and city based. Then, in these networks, it was recommended hotels to users for their next choice with link prediction methods, using the common hotels where the users prefer to stay in the past. The most successful results were obtained in the city-based network. With this study, it was tried to reduce the time spent reading the online reviews and finding the suitable hotel. To the best of our knowledge, this is the first study on recommending a hotel using online reviews by bipartite networks.

Searching Personalized $k$-wing in Large and Dynamic Bipartite Graphs

“…To explore all the edges within the same equivalence class C e , we perform BFS using e as a starting edge (line 10-17). During the exploration for the edge e, we also check for any previously explored C ′ e in e.list, if so we form the super edges between the super nodes (line [12][13][14]. Meanwhile, throughout the BFS, ∀e ∶ ψ(e) > k, the edge e stores the current super node (snID) in e.list (line 28-29).…”

“…1. The problem of personalized bipartite cohesive subgraph search can be used for a wide variety of other applications such as personalized recommendation for products [27] and hotels [13], identifying potential websites for banner advertisements [12], efficient training of the employees on the internal projects by choosing the optimal team for an employee [26] and speculating the drug-target interactions [24].…”

A Bipartite Graph-Based Recommender for Crowdfunding with Sparse Data

“…A bipartite graph divides network nodes into two types, which is different from PageRank that treats nodes as homogeneous. Only nodes in different types are directly connected, while nodes in the same type are indirectly connected [14,15]. The crowdfunding network can be abstracted as a bipartite graph, where one group of nodes is investors and the other group is items.…”

Link prediction in supernetwork: Risk perception of emergencies

“…Link prediction refers to predicting the possibility of generating a link between two nodes that have not yet generated a connecting edge in the prediction network through known information such as network nodes and the network structure [32]. The link prediction method has typical applications in terms of reconstructing networks, evaluating network evolutionary mechanisms and classifying partially labelled networks [33,34]. In recent years, in terms of the individualised recommendation system, link prediction has also been broadly applied [35].…”