GREST : a type-2 fuzzy distance model for group nearest-neighbor queries

Collaborative and group-based queries aim to find one or more points in a search space which have the minimum aggregate distance to all members of a group, situated at a set of query points. Current approaches like Group Nearest-Neighbor (GNN) queries are based on single-measure models of distance, like Euclidean distance. In reality, human has a multi-measure perception of distance so that spatial, temporal and economical aspects are important to people with possibly different individual preferences. Current approaches to GNN are unable to handle such distance measures, since it depends on the perceptions and preferences of the users. In this study, we focus on the role of users, as members of a group, situated at GNN query points. An enriched model of distance is introduced which takes the advantage of interval type-2 fuzzy sets to cope with high-order distance uncertainties, emerged from different perceptions of distance by users, and their different preferences. The flexibility of this aggregate model in handling uncertainty enables every member of the group to use a set of group-defined words to express his/her perception of multiple distance types, and to use words instead of numeric values to set the weights for each distance type according to his/her preferences. Our experimental evaluations show that the query results are closer to group preferences by providing higher quality of consensus, while keeping the spatial dispersion of the top-k results at a small level, and improved performance with reasonable response time. The proposed distance model also provides more robustness to changes of mobile member locations, eliminating unnecessary repeated computations.