A method for collaborative recommendation using knowledge integration tools and hierarchical structure of user profiles

• Profile recommendation using collective knowledge integration methods is proposed.
• Postulates for recommendation process are proposed and analyzed.
• Four integration algorithms for determining a centroid profile are proposed.
• Verification of the proposed method by experiments.

This paper proposes a new approach to collaborative profile recommendation using a hierarchical structure for user modeling. In an information retrieval system a hierarchical user profile, used to personalize the document retrieval process, is being recommended to a new user based on profiles of other, similar users. Using methodology from the Knowledge Integration domain, four criteria are defined and analyzed to complete the aim of recommendation: Reliability is required for maintaining the correct structure of the profile, O1 and O2 Optimality postulates are required to calculate the best output profile by minimizing distances to other profiles, and Conflict Solution is used to better represent situations inherent to profile recommendation. Based on those criteria, four algorithms are proposed: O1 and O2 algorithms and modified O1 and O2 algorithms. These algorithms are further analyzed to check if they provide good recommendation.