Skyline Ranking à la IR

Skyline queries have emerged as an expressive and informative tool, with minimal user input and thus, they have gained widespread attention. However, previous research works tackle the problem from an efficiency standpoint, i.e., returning the skyline as fast as possible, leaving it to the user to manually inspect the entire skyline result. Clearly, this is impractical, even with a few dozen points. The techniques addressing this issue are computationally expensive, mapping to NP-Hard problems or having exponential complexity O(2) with respect to data dimensionality d. Moreover, the result is a set, lacking any quality-based ranking. In this paper, we propose a novel IR-style ranking mechanism for skyline points, based on the renowned tf-idf weighting scheme. We present efficient algorithms to compute the quality of a skyline point according to our technique, and induce a total ordering of the skyline set. Finally, we empirically evaluate the efficiency of our method with real-life and synthetic data sets.