Assumption-Based Reasoning for Multiagent Case-Based Recommender Systems

Recommender systems (RSs) are popular tools dealing with information overload problems in eCommerce Web sites. RSs match user preferences with item representations and recommend the items that better suit these preferences. However, sometimes, the required information may not be fully available, and it could be beneficial to make conjectures about these missing values in order to generate immediately a recommendation even if not optimal. This paper presents an assumption-based multiagent RS making this type of assumptions about the user's preferences. This approach was validated in a travel application analyzing the impact of various assumption making strategies on the quality and efficiency of the recommendation process. The agents are cooperative when solving their tasks, i.e., finding the appropriate travel services (e.g., hotel or flight) to be aggregated in the final recommendation (a complete travel).

• Type: Conference
• Year: 2010
• Where: FLAIRS
• Authors: Fabiana Lorenzi, Francesco Ricci, Mara Abel, Ana L. C. Bazzan
• Keywords: Artificial Intelligence, Assumption-based Multiagent Rs, FLAIRS 2010, Information Overload Problems, Match User Preferences