Finding Superior Skyline Points from Incomplete Data

The skyline query has proven to be an important tool in multi-criteria decision making and search space pruning. A skyline query returns the subset of points from a multidimensional dataset that are not dominated by any other point. Due to its wide applications, skyline query and its variants have been extensively studied in the past. However, skyline computation for incomplete domain, where points have missing values for some dimensions, has not received enough attention. The existing solutions for such incomplete datasets use weak pareto dominance relation which is nontransitive and cyclic. Hence, many of the desirable points are not included in the skyline. Consequently, the skyline no longer offers a reliable overview of the dataset. Moreover, the skyline set returned by these methods is unordered and has high cardinality. The end user does not have control over the result size. Therefore, we have adapted the top-k frequent skyline approach proposed for complete datasets to find interesting points from incomplete datasets. The proposed approach overcomes the above mentioned drawbacks and returns top-k points ordered by their fractional skyline frequency. Experimental results on both synthetic and real world datasets demonstrate the ability of our approach to find superior skyline points from incomplete datasets.