Top-k ranking for uncertain data

The goal of top-k ranking is to rank individuals so that the best k of them can be determined. The definition of top-k ranking is easy for certain data. But for uncertain data, the problem becomes challenging, both semantically and computationally. In this paper, we study semantic issues with top-k ranking for objects modeled by uncertain data in databases. Uncertain data of objects have different formats such as probability distribution of the values of objects or relations among the values of objects. We present a ranking theory so that uncertain data of objects with different formats can be reasonably employed to define the top-k objects. We first define this theory using possible world semantics for discrete data. Then we give the definition in high-dimension space so that it can handle both discrete and continuous data. We further extend this theory to consider weights of positions in top-k ranking.