Abstract

In many data mining applications, the k nearest neighbor join is widely utilized as a primitive operation. It is a combination of the k nearest neighbor query and the join operation. In recent years, the k nearest neighbor join in high dimensional data is a topic of interest for researchers. By the way of the Hilbert curve, the points in a multi-dimensional space can be mapped into one dimension. However the Hilbert curve can not always preserve the spatial locality. Sometimes the potential errors may exit. The randomly shifted versions of the input data set are produced independently. The kNN join can be operated for each randomly shifted copy. Then better results will be achieved.

Paper Details