Skyline Queries with Noisy Comparisons

We study in this paper the computation of skyline queries - a popular tool for multicriteria data analysis - in the presence of noisy input. Motivated by crowdsourcing applications, we present the first algorithms for skyline evaluation in a computation model where the input data items can only be compared through noisy comparisons. In this model comparisons may return wrong answers with some probability, and confidence can be increased through independent repetitions of a comparison. Our goal is to minimize the number of comparisons required for computing or verifying a candidate skyline, while returning the correct answer with high probability. We design output-sensitive algorithms, namely algorithms that take advantage of the potentially small size of the skyline, and analyze the number of comparison rounds of our solutions. We also consider the problem of predicting the most likely skyline given some partial information in the form of noisy comparisons, and show that optimal pre...

• Type: Journal
• Year: 2015
• Where: PODS
• Authors: Benoît Groz, Tova Milo
• Keywords: Database, PODS 2015