A distributed recommender agent model based on user's perspective SVD technique

• Main Authors: Praserttitipong,D., Sophatsathit,P.
• Format: Article
• Published: Advanced Institute of Convergence Information Technology Research Center 2015
• Subjects: Computer Networks and Communications; Software
• Online Access: http://www.scopus.com/inward/record.url?partnerID=HzOxMe3b&scp=84862727744&origin=inward; http://cmuir.cmu.ac.th/handle/6653943832/38633
• Institution: Chiang Mai University

This paper proposes a modified centralized recommender technique to be embedded in a software agent. The software agent will act as a cooperative assistant to carry out some tasks on behalf of its user working in the decentralized environment. The historical rating data from other available sources and itself are learned so as to constructing the agent's prior knowledge at the central server. Knowledge acquisition is accomplished by an eager learning process based on adaptive user's perspective singular value decomposition (USVD) technique. Subsequent incremental knowledge update according to user's feedback is maintained by a lazy learning process until some thresholds are reached. Thereby the eager learning process takes over. The overall accumulated knowledge entails the agent to arrive at more accurate recommendations. The processing time for this knowledge maintenance is reduced from polynomial time to constant time. Moreover, the versatility of the proposed model renders it to be applied to sparse knowledge and new-item cold-start problems. This software agent is referred to as a recommender agent.