Aggregation of data quality metrics using the Choquet integral
Résumé
In the context of multi-source databases, data fusion is a tricky task, and resolving inconsistency problems when merging duplicate information is one of the most intricate issues as it is generally resolved through subjective approaches. Using data quality dimensions may help sort out such a question impartially. Quality metrics are the objective criteria that justify the preference of a value v1 over a value v2; where v1 and v2 are both referring to the same real world entity but issue from different sources. However, this technique is fairly complicated when the v1 quality criteria are not all better than the v2 ones; when we have to choose, for instance, between a highly fresh but inconsistent data, and a consistent old one. Hence, we need a global qualifying score to facilitate the comparison. In this perspective, aggregation of data quality metrics can be the solution for computing a global and objective data quality score. In this paper, we introduce a solution that uses the Choquet integral as a means of aggregating data quality metrics.