Abstract
A group package recommender framework is proposed to provide recommendations on dynamically defined packages of products and services to large heterogeneous groups based on multi-criteria optimization. The framework is based on: (1) sampling the entire large group, (2) eliciting the utility function for each member, (3) clustering the sample heterogeneous group into a number of relatively small homogeneous subgroups, (4) extracting the representative utility function for each subgroup, (5) estimating the utility function of the entire group, and use it to find an optimal recommendation alternative, (6) diversify recommendations across those subgroups, (7) applying a group decision-making method, from social choice theories, to refine the recommendations. A preliminary experimental study is conducted, which shows that the proposed framework is able to produce a small set of ranked recommendations that retains close to optimal precision and recall, as compared to the baseline method applied directly to original large groups.