A mobile 3D-GIS hybrid recommender system for tourism
Introduction
In the last decade, the growth of the Internet and its users has heavily modified several social behaviors in many fields. In tourism the web has originated the electronic tourism (e-tourism) [75], [76]. According to recent studies (e.g. European Travel Commission), the web is the primary source of information for people in Western countries for planning trips and obtaining information about travel destinations [33]. Despite this fact, most tourism web sites only offer basic booking functionality, and are unable to support customers in their searches. Clearly, complexity and quantity of e-tourism products might become a problem because users cannot look up entire databases in order to choose the most suitable products for them. Therefore, frustrated users will likely give up the web site without any purchase.
Personalization [24] has recently been recognized by experts like a critical factor of e-tourism companies to be successful. Different personalization techniques have emerged to deal with this problem. The most successful one has been the recommender systems (RSs) [2], [59], [66], which lead users to suitable products by using different information filtering techniques that utilize knowledge about their personal preferences. The most popular approaches in e-tourism personalization are collaborative filtering and content based techniques [23], [60].
Recently, we proposed REJA [46], [63] a web-based restaurant georeferenced RS for the province of Jaén in southern Spain. REJA provides basic personalization services to those tourists that visit Jaén paying attention to the cold start problem [4], [74] and overcoming it by a hybrid technique based on a knowledge based approach. However, new challenges in the e-tourism industry require these services to be available without spatial–temporal limitations and provide context-aware information [22].
These new e-tourism challenges have been tackled by using mobile computing (e.g. mobile phones, tablets, personal digital assistants, etc.) [23], [35], [61], [67]. Because such platforms may add ubiquity providing services to costumers at anytime and anywhere; connectivity with wireless networks (e.g., GPRS and UMTS) at a relatively low cost [18]; location-awareness by means of Global Positioning Systems (GPSs) and compasses that enable to take contextual knowledge (physical location, motion speed, time, etc.) of users into consideration [3], [23]; and graphics capabilities because recent high-end mobile devices include Graphics Processing Units (GPUs) and large display sizes, which open the door to the development of 3D user interfaces [5].
Context-Aware Recommender Systems (CARSs) are a special type of RS that takes users’ contextual knowledge into account when providing recommendations. Early CARS designed for mobile devices mostly rely on the user’s location to determine the appropriateness of POIs to be provided to the customers, see for example [1], [19], [34], [56], [69], [73]. However more advanced solutions are capable of taking advantage of customer’s previous choices information by means of content-based filtering [39], [65], [77] or exploiting knowledge about other customers with similar interests by means of collaborative filtering [10], [31], [32], [36].
Apart from a recommender engine, the success of a mobile tourism support system depends on an intuitive and usable interface. So far, because of the limited size of mobile devices, user interaction is especially difficult and frustrating. All reported solutions portray recommended items within a textual interface or, at best, within an abstract 2D map [35]. However, both approaches either present severe usability problems or require a cognitive effort to understand the provided information.
Taking into account the previous features and drawbacks presented by current mobile e-tourism personalized systems, we present different extensions upon our system REJA in order to address different aspects related to mobile recommender systems and overcome interface drawbacks. First, we extend our system to include user’s location, such that, the recommendations are restricted to an influence area determined by the user’s location together with her preferences and being generated by a hybrid recommender approach. Second, we provide an interface for georeferencing recommendations with a 3D view of the terrain with actual imagery (satellite/aerial) and landmarks that would be easier and faster to understand [52] by using a solution integrated into a mobile 3D GIS. This allows users to benefit from innovative features including 3D geovisualization, progressive transmission of terrain across wireless networks, geolocation and possibility to download location-sensitive recommendations to their mobile devices on demand. Third, this paper describes and discusses a complete client–server architecture which implements these ideas. Eventually, an evaluation survey about the system usability and its performance is presented.
We must highlight that contrarily to previous solutions that investigated mobile recommender systems by using a desktop-based system simulating a mobile device (e.g., [39], [77]), we have carried out our implementation and evaluation on real smartphones. Therefore, the proposed algorithms and strategies have been designed aiming at efficiency, simplicity and usability on small display sizes.
The rest of the paper is organized as follows. Section 2 provides a necessary background about recommender and geographic information systems. Section 3 presents our proposal of a georeferenced hybrid recommender system designed for mobile tourism. In order to validate this proposal, we have built a fully operational prototype. Section 4 introduces this prototype and outlines its design and architecture. Section 5 presents a survey about the usability and performance of the proposed system. Finally, Section 6 concludes the paper.
Section snippets
Preliminaries
This Section provides a brief review about recommender systems describing the most common models, paying then attention to the main features and problems of the systems used by our proposal. Following, due to the fact that our system makes a heavy use of GIS and 3D mobile technologies, some general concepts are provided.
A 3D-GIS mobile location-aware recommender system
Here a 3D context-aware mobile recommender system is introduced. This proposal could be applied to different e-tourism purposes, but in this paper we upgrade the REJA system [46], [63] from a web-based RS to a 3D map-based mobile CARS whose goals are:
- •
Ubiquity: users may use the system wherever they like. The mobile platform and GIS are key factors to achieve this goal.
- •
Location-awareness: because of the ubiquitous nature of the system, it seems necessary and convenient that the recommendations
Architecture design and implementation
We have developed a fully operational mobile recommender system that implements the proposed recommendation approach. This Section introduces our prototype whose goal is to provide recommendations to users about restaurants in the province of Jaén, Spain. This Section also describes the architecture and discusses issues related to the design and usage of our specific implementation.
We propose a three-tier client–server architecture, as illustrated in Fig. 7. It consists of three software
Evaluation
Evaluation of RS is usually performed using offline experiments and measured in terms of precision and accuracy in the recommendation [70]. However, this type of evaluation presents some limitations [8], for example, contextual conditions are difficult to emulate in an offline environment. Therefore, and following similar works on CARS and tourist guides for mobile devices [8], [25], [62], [65], we have conducted a user study to evaluate our system.
We have built a fully operational prototype
Concluding remarks
The new generation of smartphones and tablets include features which were unconceivable some years ago: surprising graphics capabilities, built-in sensors and ubiquitous connection to the Internet. This paper presents a novel recommender system that takes full advantage of these features. Our proposal consists of integrating a location-sensitive hybrid recommender engine with a custom-made 3D GIS architecture capable of running interactively on modern mobile devices.
Our solution adapts the
Acknowledgments
This work has been partially supported by the Ministerio de Ciencia e Innovación (Spain) and the European Union (via ERDF funds) through the research projects P07-TIC-02773 and TIN2009-08286.
References (77)
A new similarity measure for collaborative filtering to alleviate the new user cold-starting problem
Information Sciences
(2008)- et al.
Exploring synergies beween content-based filtering and spreading activation techniques in knowledge-based recommender systems
Information Sciences
(2011) - et al.
Collaborative filtering based on significances
Information Sciences
(2012) - et al.
Navigation techniques for small-screen devices: an evaluation on maps and web pages
International Journal of Human–Computer Studies
(2008) - et al.
Building and evaluating a location-based service recommendation system with a preference adjustment mechanism
Expert Systems with Applications
(2009) - et al.
A hybrid of sequential rules and collaborative filtering for product recommendation
Information Sciences
(2009) - et al.
Personalized recommendation of popular blog articles
Information Sciences
(2011) - et al.
Navigating large terrains using commodity mobile devices
Computers & Geosciences
(2011) - et al.
A hybrid recommender system for the selective dissemination of research resources in a technology transfer office
Information Sciences
(2012) - et al.
A google wave-based fuzzy recommender system to disseminate information in university digital libraries 2.0
Information Sciences
(2011)