DESIGN AND IMPLEMENTATION OF A MOBILE RECOMMENDER SYSTEM FOR TOURISTS

Abstract

Mobile tourism recommender systems have become a promising and interesting research direction in recent years, as a result of the vast use of mobile devices in our day-to-day lives, and also the increasing desire by humans for travel and adventure. Many system designs and approaches have been proposed in this direction. However, with each approach comes the need to face the key challenge of recommender systems - the cold-start situation. In this thesis, a switching feature based model that leverages the need of both the new and existing users is proposed. At the point of entry, the system checks to ascertain if the user is a new or an existing user. For a new user, the system collects demographic data (on budget size of user, preferred location size and preferred cultural affiliation) that matches the users’ needs and stores it for future reference or passes it to the Bayesian algorithm from which recommendations are obtained. The user can change location type/demographic data and obtain further recommendations, or exit the system. For an existing user, the system switches to the collaborative-filtering subsystem, where the user inputs the location facility he wants. This data is passed to the collaborative algorithm, and recommendations are obtained. Also, at this point, the user can either change location type for other recommendations or exit the system. Recommendation results are produced through the appropriate algorithm and offered based on the items in the database. The results of evaluation using discounted cumulative gain, precision, recall, and mean absolute showed the system is efficient with less errors in its recommendations Also, experimental results obtained through questionnaire distributed to general users, users from the computer science domain as well as experts in the tourism domain, showed the effectiveness of the proposed techniques.