Toward Geospatial Collaborative Tourism Recommender Systems

1. Introduction

Tourism as a social, cultural, and economic phenomenon entails the movement of people to countries or places outside their living environment. Tourism activities are highly dependent on spatial information such as places of interest and attractions that are likely to present an interest for potential visitors. Planning a trip and finding the most appropriate travel destinations and Points of Interests (POIs) are still considered as open research issues in GIScience research applied to the tourism domain. A POI can represent a specific tourist attraction that a user may find interesting to visit such as a museum, historical and cultural buildings. Nowadays, the Internet and the World Wide Web offer huge amounts of free information on travel destinations that may be useful for users wishing to plan a visit. However, the large extent of information available is often overwhelming and time consuming for most users as selecting the most appropriate POIs and planning a tourism trip according to their own preferences is far from being a straightforward task. In order to overcome this information overload problem, Tourism Recommender Systems (TRSs) have been recently introduced in order to suggest items that are most likely of interest to a given user. The main principles behind TRSs are that items of interest should be identified and recommended by taking into account the user’s profile and preferences. In other word, TRSs infer users’ needs by considering either explicit or implicitly given users’ profiles and even by comparing such profiles to users with similar preferences.

This chapter provides a survey of TRSs, its main objectives are as follows:

• 1.

  To briefly review current tourism recommendation approaches as well as pros and cons. We introduce different classes of tourism recommendation approaches including from content-based, collaborative, knowledge-based, demographic, to hybrid TRSs. Content-based TRSs suggest items that are similar to the ones that the user liked before, while collaborative TRSs take into account items that similar users liked in the past. Knowledge-based TRSs suggest items based on specific domain knowledge and user's preferences. Demographic TRSs suggest items based on some basic user demographic. Finally, hybrid TRSs take into account the combination of the above mentioned techniques.

• 2.

  To review current approaches oriented to the representation of user profiles in TRSs. We will introduce user profile representations such as the ones based on a list of keywords (i.e., as categories associated to the users), vectors (i.e., where each vector value denotes a degree of interest associated to a given item), semantic models (i.e., based on domain-based ontologies) and uncertainty-based models (i.e., to take into account the uncertainty associated to the information stored in the user profile).

• 3.

  To present the different functionalities so far provided by current TRS. Such functions are classified according to the services delivered such as personal interest estimation, travel destination and tourist packs recommendation, ranked list of POIs recommendation, route or itinerary, recommendations based on Location-Based Social Networks (LBSN), and group-based recommendations.

• 4.

  To provide an insight regarding some additional influential factors that play a role in TRSs including geographical, temporal and social influences.

• 5.

  To introduce the different Computational Intelligence (CI) techniques applied in TRSs including statistics, fuzzy sets, artificial neural networks, evolutionary computing, and swarm computing.

• 6.

  To provide an insight to the different evaluation metrics currently applied to TRSs including quantitative and qualitative metrics either explicit or implicit.

• 7.

  To introduce a case study that shows the potential of TRS. The experiment developed uses Foursquare LBSN data including both their geographical and social dimensions. The objective is to take into account several factors such as social relations and similarities while making some choices. The objective is to recommend some optimal routes that take into account some recommended POIs. The whole approach is quantitatively evaluated. The implementation presents some preliminary findings and future works.