Multi- agents integrated mobile services for travellers

  • Hend Al-Tair

Student thesis: Master's Thesis


Hend Abdulrazzaq Al Tair. Multi-Agents Integrated Mobile Services for Travellers. M.Sc. By Research Thesis, Computer Engineering Department, Khalifa University of Science, Technology and Research, Sharjah, United Arab Emirates, December 2011 This thesis investigates the different recommender systems approaches and their filtering techniques. It tackles different possible solutions to extend two-dimensional recommender systems to be multi-dimensional through using additional contextual information which makes the services provided to users more personalized. Additionally, it highlights the pro-activity aspect of recommender systems which can be achieved by integrating agent technology with the recommender system. The system proposed in this thesis integrates a Multi-Dimensional (MD) rating approach with multiagents to obtain the appropriate recommendations for the user. The MD approach uses information about the user and other contextual information. It is based on extending a knowledge-base hybrid recommender system with reduction-based theory. The prediction and the rating process by the recommender system use the conditional probability with multi-attribute theory. The inference engine of the system uses agents which are specialized in different services to serve users. This approach results in a system that is pro-active. The main distinctive features of the proposed system are:  Pro-activity: The system checks the available information and takes the initiative of suggesting various services to the user before they even get requested.  Builds users' profiles from scratch: The system provides services to users who did not provide any profile information. Based on the history of users' selections of the recommended services the system can build the profiles in the background. Consequently, the effectiveness of the system was evaluated by real users who performed over 200 test cases with different profiles under a variety of scenarios. Overall the system performed well and the results indicate that the system is capable of dealing with users with varied degrees of profiles and is able to deliver suitable services.
Date of Award2011
Original languageAmerican English
SupervisorMohamed Zemerly (Supervisor)


  • Recommender System
  • Multi-agent
  • MD
  • Rule-based
  • Pro-activity

Cite this