Preference Mining in the Travel Domain

Abstract

Personalization of user experience through recommendations involves understanding their preferences and the context they are living in. In this work, we present a method to rank travel offers returned in response to a travel request made by a user. To give a sensible answer, we learn users' preferences over time and use them to understand travelers' needs. Our solution is based on a data-mining-based recommender system. We first design a database of historical traveler data and populate it with data generated according to rules mimicking the features of actual user profiles. These rules are then used as ground truth to validate the accuracy of the proposed learning algorithm. After performing data pre-processing, a knowledge base is set up by mining association rules from the database, which will then be used along with the travel request to assign a score to each of the potential travel offers, thus ranking them. To test the proposed methodology, we generate synthesized data according to some distributions. The results of the experiments approve the effectiveness of the proposed ranking mechanisms. Finally, we demonstrate the presentation of the ranked offers to the user via some mock-ups of the intended application.

THOR: A Hybrid Recommender System for the Personalized Travel Experience

“…Javadian et al [39] propose a data-mining-based RS to rank offers. They use association rule mining to calculate the similarities between the user requests and the offers.…”

“…In our work, we design and build the RS block proposed by Javadian et al [8,9]. To do so, we combine collaborative and content-based methods to develop a hybrid approach.…”

“…Recently, there have been various advancements in the field of Context-Aware Recommender Systems (CARS) [5][6][7]. Indeed, one should also take into account that travel preferences may be significantly influenced by the context in which a traveler interacts with the system [8,9]. According to Dey [10], "Context is any information that can be used to characterize the situation of an entity, where an entity can be a person, place, or physical or computational object" and "A system is context-aware if it uses context to provide relevant information and services to the user, where relevancy depends on the user's task".…”

“…In this work, we assume that the user has access to an application to ask for travel offers, such as, for example, the Travel Companion (TC) provided by the Shift2Rail ecosystem (see Section 2.1). To provide a personalized experience to travelers, Javadian et al designed a high-level system architecture in [9] and a contextual preference model in [8] using the Context Dimension Tree methodology [17]. This work aims to implement the recommender core of the TC (or of a TC-like application) based on the contextual preference model presented in [9].…”