•We address the problem of simultaneous computation of skyline probabilities of multiple objects.•Our method is based on a novel concept of zero-contributing set and multi-level prefix-based absorption.•We propose a constraint-based k-level absorption to identify zero-contributing sets related to m objects at a time.•One of the major design issues is to determine the number m of reference objects. We also analyse the choice of m.•Detailed experimental analysis for real and synthetic datasets are reported to corroborate the efficiency of our algorithm. The problem of recommending objects based on attributes is a novel recommendation problem. When the preferences of attributes are uncertain and are expressed in terms of probabilities, the recommendation problem boils down to computing skyline probabilities of all objects in the database. Though there exists efficient algorithms to compute skyline probability of a single object when pair-wise preference probabilities are given, the problem of computing skyline probabilities of all objects in the database is not yet solved. In this paper, we establish the concept of preference probability over uncertain preferences in the context of a recommender system. We propose an efficient approach to address the problem of simultaneous computation of skyline probabilities of multiple objects. Our method is based on a novel concept of zero-contributing set and multi-level prefix-based absorption. The idea is to carry out the absorption with multiple reference objects. One of the major design issues is to determine the number of m reference objects. We also analyse the choice of m. We report extensive experimental analysis to justify the efficiency of our algorithm.