K-nearest neighbor search for fuzzy objects

Abstract

The K-Nearest Neighbor search (kNN) problem has been investigated extensively in the past due to its broad range of applications. In this paper we study this problem in the context of fuzzy objects that have indeterministic boundaries. Fuzzy objects play an important role in many areas, such as biomedical image databases and GIS. Existing research on fuzzy objects mainly focuses on modelling basic fuzzy object types and operations, leaving the processing of more advanced queries such as kNN query untouched. In this paper, we propose two new kinds of kNN queries for fuzzy objects, Ad-hoc kNN query (AKNN) and Range kNN query (RKNN), to find the k nearest objects qualifying at a probability threshold or within a probability range. For efficient AKNN query processing, we optimize the basic best-first search algorithm by deriving more accurate approximations for the distance function between fuzzy objects and the query object. To improve the performance of RKNN search, effective pruning rules are developed to significantly reduce the search space and further speed up the candidate refinement process. The efficiency of our proposed algorithms as well as the optimization techniques are verified with an extensive set of experiments using both synthetic and real datasets.

Spatial Co-Location Pattern Discovery from Fuzzy Objects

“…Based upon that, more advanced spatial queries such as the nearest neighbor query were discussed in Ref. 21. However, it remains largely untouched to answer the spatial data mining problem such as co-location pattern mining, which is addressed in this paper.…”

“…However, in some real applications such as biomedical image analysis and geographical information systems (GIS), the data may not satisfy this assumption. 21 In real life, such as "old man", "tall tree", etc, the boundary of these objects cannot be identified easily, these objects are known as fuzzy objects. Although fuzzy objects have long been studied in GIS community, 23,24 spatial co-location patterns mining still remain uninvestigated.…”

Constrained k Nearest Neighbor Query for Uncertain Object in the Network

“…Assumed fuzzy object set D= {O 1 For the fuzzy object O 1 , its 0.5-cut set is O 1/0.5 = {o 12 , o 13 }. The 0.5-cut set of the fuzzy object O 3 is O 3/0.5 = {o 32 , o 33 }.…”

Top-k Similarity Join over Multi-valued Objects

“…Many variation of join queries over multi-dimensional space have been studied in different contexts, including road networks [18] and moving objects [20]. Spatial queries such as nearest neighbor queries over fuzzy objects have been recently studied [22]. Fuzzy objects possess similar semantics as uncertain objects (e.g., instances are mutually exclusive).…”

“…Fuzzy objects possess similar semantics as uncertain objects (e.g., instances are mutually exclusive). The techniques in [22] are not applicable to the problem studied in our paper due to the different semantics as well as inherent difference in query types.…”