Probability threshold based group nearest neighbor queries for fuzzy objects

Abstract

Fuzzy objects have diversi ed applications in various promising areas like biomedical image pro- cessing and Geographical Information System (GIS). Existing work on fuzzy objects mainly indulge on modeling the basic fuzzy objects and processing of a simple query i.e., nearest neighbor query on those objects. However, the processing of more advanced queries such as group nearest neighbor, reverse nearest neighbor, reverse nearest neighbor join, and skyline queries are still remain unex- plored in the literature. Therefore, in this thesis, we explore the problem of evaluating the group nearest neighbor queries for fuzzy objects based on user speci ed probability threshold. Hence, we introduce two new kinds of group nearest neighbor queries considering a single probability thresh- old and a probability threshold interval for fuzzy objects namely, fuzzy group nearest neighbor (FGNN) query and continuous fuzzy group nearest neighbor (CFGNN) query respectively. Given a set of fuzzy data objects, and a group of fuzzy query objects, a FGNN query and a CFGNN query will retrieve the fuzzy objects that minimize the aggregate distance (e.g., MAX, and SUM) to the group at a probability threshold and within a probability interval respectively. Existing algorithms for group nearest neighbor queries for point objects cannot be directly deployed for our FGNN and CFGNN queries. Therefore, we propose e cient algorithms with e ective pruning conditions to compute the group nearest neighbors for fuzzy objects. An extensive experimental study depicts the e cacy of our proposed techniques.