Optimizing the Performance and Robustness of Type-2 Fuzzy Group Nearest-Neighbor Queries

Abstract

In Group Nearest-Neighbor (GNN) queries, the goal is to find one or more points of interest with minimum sum of distance to the current location of mobile users. The classic forms of GNN use Euclidean distance measure which is not sufficient to capture other essential distance perceptions of human and the inherent uncertainty of it. To overcome this problem, an improved distance model can be used which is based on a richer, closer to real-world type-2 fuzzy logic distance model. However, large search spaces as well as the need for higher-order uncertainty management will increase the response times of such GNN queries. In this paper two fuzzy clustering methods combined with spatial tessellation are exploited to reduce the search space. Extensive evaluation of the proposed method shows improved response times compared to naïve method while maintaining a high quality of approximation. The proposed uncertainty management method also provides robustness to movement of mobile users, eliminating the need for full re-computation of candidate clusters when the locations of group members are changed.