A machine learning algorithm for identifying inconsistencies among sets of restrictions
In the European Central Bank (ECB) a new Centralised Submission Platform has been developed which supports the collection of any kind of data types and formats from reporting agents. Upon file reception, data quality checks are applied instantly to the submitted data. These checks are represented in the form of assertions at the platform’s validation engine. We developed a machine learning program that, regardless of the specific dataset, (i) it identifies logical and computational inconsistencies in the business validation checks and (ii) generates synthetic data satisfying the underlying statistical or economic models.