Welcome to the MAGReS wiki!

MAGReS (Multi-Agent Group Recommender System), consists of a MAS in which each agent represents a group member (or user within the group). MAGReS, initially proposed for group recommendation of movies in (Villavicencio et al., 2016), uses a similar idea than the approaches based on recommendation aggregation, but instead of using of regular aggregation techniques (such as average or least misery) it uses a MAS to select those individual recommendations that will be part of the group recommendations.

Approach

For example, let us assume that we want to generate recommendations in the domain of LBSN, and that we have a group of three friends who want to go to a restaurant together and a set of P possible restaurants to be chosen. According to the main idea of MAGReS, each user has his/her own personal agent that is able to access the user profile. For simplicity, a profile includes only ratings over (a subset of) the possible restaurants. A user rating rt_i(item) is a value (in the range [0,1] where 0 means dissatisfaction and 1 means high satisfaction) assigned by the user i to the given POI or restaurant.

In this context, let us consider the following (initial) situation:

• ag_{1} manages ratings <rt_{1}(POI1)=0.6, rt_{1}(POI2)=0.8> for user #1,
• • ag_{2}manages <rt_{2}(POI1)=0.4, rt_{3}(POI3)=0.6>for user #2, and
• • ag_{3} manages < rt_{3}(POI2)=0.2, rt_{3}(POI3)=0.8> for user #3.

Then we have A={ ag_{1}, ag_{2}, ag_{3}} as the MAS in which the negotiation for the “best” POI or restaurant to visit (the one that will satisfy all the three agents) takes place. Also, these users are friends and, thus, they are related in a social network. They move around the areas shown in the figure.