Research on Uncertain Data Provenance Based on Confidence Computing

摘要

Data provenance describes data generation process and evolves with time. Most of the existing work has so far aimed at certain data; however, many applications (especially super-large scale applications) are related to uncertain data, such as economy, logistics, finance, telecom. In the course of evolution, the uncertainty is transferred and magnified in each cycle, and has an enormous influence on the quality of query results. A concept of minimal how-provenance is proposed in this paper and it proved to be a kind of provenance semiring. In addition, it can be used to calculus the confidences of the results.