Error Correction for Massive Data Sets

The paper is concerned with the problem of automatic detection and correction of errors into massive datasets. As customary, erroneous data records are detected by formulating a set of rules. Such rules are here encoded into linear inequalities. This allows to check the set of rules for inconsistencies and redundancies by using a polyhedral mathematics approach. Moreover, it allows to correct erroneous data records by introducing the minimum changes through an integer linear programming approach. Results of a particularization of the proposed procedure to a real-world case of census data correction are reported.