M-Skyline: Taking sunk cost and alternative recommendation in consideration for skyline query on uncertain data.

Traditional probabilistic skyline query over uncertain data returns a tuple of individual recommendations for customers. However, the uncertainty of the dataset brings the possibility that the recommendation is not correct. Once the incorrect candidate is recommended, user needs to query the skyline again (may use a higher probability threshold) and tries to find alternatives. This greatly hurts user experience for those recommendation scenarios where finding out query results to be wrong brings non-negligible sunk cost, such as spending time to visit a recommended interest point. To address this concern, we propose a novel M-Skyline query model that takes consideration of sunk cost and offers backup recommendation. Moreover, in order to optimize the query speed for finding such M-Skyline results, we devise several fast query algorithms. Extensive experiments with both real and synthetic datasets demonstrate the effectiveness and efficiency of our proposed algorithms under various scenarios.