Discovering strong skyline points in high dimensional spaces

Abstract

Current interests in skyline computation arise due to their relation to preference queries. Since it is guaraneed that a skyline point will not lose out in all dimensions when compared to any other point in the data set, this means that for each skyline point, there exists a set of weight assignments to the dimensions such that the point will become the top user preference. We believe that the usefulness of skyline points is not limited to such application and can be extended to data analysis and knowledge discovery as well. However, since the skyline of high dimensional datasets (which are common in data analysis applications) can contain too many points, various means must be developed to filter off the less interesting skyline points in high dimensions. In this paper, we will propose algorithms to find a set of interesting skyline points called strong skyline points. Extensive experiments show that our proposal is both effective and efficient.