A probabilistic decision model for entity matching in heterogeneous databases

In recent years, there has been a proliferation of database systems in all types of organizations. In many cases, these databases are developed in different departments and maintained autonomously. Much is to be gained, however, if databases across departments, divisions, or even organizations can be related to one another. One main problem of relating data stored in different databases is the differences in their representation of real-world entities, such as the use of different identifiers or primary keys. We present a decision theoretic model for matching entities across different databases. The decision to match two entities from two different databases inherently involves some uncertainty since an exact match may not be found because of errors in data collection, data entry, and data representation. We model this uncertainty using probability theory and propose an integer programming formulation that minimizes the total cost associated with the entity matching decision. The model has been implemented and validated on real-world data.