K-dominant skyline queries on incomplete data

Abstract

The skyline query has been extensively explored as one of popular techniques to filter uninteresting data objects, which plays an important role in many real-life applications such as multi-criteria decision making and personalized services. This query has also been incorporated into commercial database systems for supporting preference queries. However, a skyline query may retrieve too many objects to analyze intensively especially for high-dimensional datasets. As a result, k-dominant skyline query has been introduced to control the number of the objects retrieved. Existing algorithms for k-dominant skyline queries only aim at complete data, which is not well-suited for incomplete data, even though incomplete data is pervasive in scientific research and real life, due to delivery failure, no power of battery, accidental loss, etc. In this paper, we systematically study the problem of k-dominant skyline queries on incomplete data (IkDS), where the data objects might miss their attribute values. We formalize the IkDS query and then present three efficient algorithms for finding k-dominant skyline objects over incomplete data. Several novel concepts/techniques are utilized including local skyline, dominance ability, and bitmap index on incomplete data to shrink the search space. In addition, we extend our techniques to tackle two interesting variants, i.e., weighted dominant skyline query and top-δ dominant skyline query, over incomplete data. Extensive experiments using both real and synthetic data sets demonstrate the performance of our proposed algorithms.