Fast Probabilistic Ranking under x-Relation Model

TL;DR

This paper introduces a novel algorithm for probabilistic top-k queries under the x-Relation model that reduces computation time from quadratic to linear in dataset size, improving efficiency in uncertain data ranking.

Contribution

The paper presents a new algorithm that computes tuple ranking probabilities in O(kn) time, significantly enhancing efficiency over previous O(kn^2) methods.

Findings

Reduces probability computation complexity from O(kn^2) to O(kn)

Improves efficiency of probabilistic top-k queries in uncertain data

Applicable to x-Relation model for probabilistic ranking

Abstract

The probabilistic top-k queries based on the interplay of score and probability, under the possible worlds semantic, become an important research issue that considers both score and uncertainty on the same basis. In the literature, many different probabilistic top-k queries are proposed. Almost all of them need to compute the probability of a tuple t_i to be ranked at the j-th position across the entire set of possible worlds. The cost of such computing is the dominant cost and is known as O(kn^2), where n is the size of dataset. In this paper, we propose a new novel algorithm that computes such probability in O(kn).