| Abstrakt: |
The Web has become the primary source of information for people when searching for suitable travel products before traveling. The explosive growth, variety of information available on the Web, and the rapid development of new e-business services (such as buying products and product comparisons) frequently overwhelm users, leading them to make poor decisions. Recommender systems have been proven to be a promising solution to the problem of information overload. Based on a review of overseas literature during the past ten years, we discussed the concept, application, and latest research progress of travel recommender systems. We differentiate the concepts of recommender systems and information retrieval systems (e.g., search engines) based on the criteria of individualized, interesting, and useful. Information retrieval systems often offer the same results for each request while recommender systems will consider personal needs and present different results. Some travel recommender systems such as Trip@dvice (ECTRL), tripmatcher (triplehop), and MePrint (VacationCoach) are widely used by destination organizations or e-business companies. To provide some references for the research on travel recommender systems, we ranked journals found in the Springer, ScienceDirect, EBSCO (Hospitality & Tourism Complete), and IEEE Explore databases from 1999 to 2013. Some famous academic conferences are recommended as well. We classified travel recommender systems according to the recommended technology, items, and devices used by recommender systems. Focusing on the key technologies applied in tourism recommender systems, the complexity and particularity were analyzed in the tourism and travel industry. Due to the limitations of traditional methods, such as collaborative filtering and content-based filtering, knowledge-based filtering and hybrid methods were more adopted for travel recommender systems. Some applications of travel decision theory--such as the travel destination choice model, travel decision style, and the behavior framework for destination recommendation systems design--were discussed to demonstrate the theory's importance in designing a good travel recommender system. A general framework for travel recommender systems which includes user profiling, recommend computing, and results presentation is presented. When designing a travel recommender system, the following factors should be considered: (1) The system can recommend a bundling of elementary components rather than a single destination or product. (2) Both short-term and long-term preferences must influence the recommendation. Short-term preferences should have greater weight than long-term preferences. (3) The cognitive effort that the user devotes to the information search should be reduced. More implicit methods should be used to elicit the user's preferences. (4) System bootstrapping without an initial memory of rating interactions should be allowed (Unregistered users can also receive useful recommendations). (5) Human/computer interaction such as asking/answering conversational mode or proposing/criticizing conversational mode should be supported. Future studies should focus on the following aspects: (1) More information and skills should be used for comprehensive understanding of users and items. (2) More contextual information should be considered to extend the traditional two-dimensional User x Item space. (3) Recommender systems will incorporate more multi-criteria rating information into the recommendation process to improve the quality of recommendations by providing additional information and being able to represent more complex preferences of each user. (4) With the development of 3G and cloud computing technology, mobile recommender systems will be a promising area in the tourism and travel industry. Information from social networks can be integrated into recommender systems to produce more accurate recommendations. (5) User privacy protection is also a challenge for recommender systems as they often need as much personal information as possible. So methodologies for protecting user anonymity and privacy are required, and they should guarantee the effectiveness and accuracy of recommendations without compromising the privacy of user profiles and sensitive contextual information. [ABSTRACT FROM AUTHOR] |
