A systematic approach to reliability assessment in integrated databases

摘要

We provide a novel framework based on a systematic treatment of data inconsistency and the related concept of data reliability in integrated databases. Our main contribution is the formalization of reliability assessment for historical data where redundancy and inconsistency are common. We discover data inconsistency through the analysis of relationships between existing reports in the integrated database. We present a new approach by defining properties (rules) that a good measure of reliability should satisfy. We then propose such measures and show which properties they satisfy. We also report on a simulation-based study of the introduced framework.