Deciding Confidence for Identity Resolution in Closed and Open Universes of Entity Identity Information

Entity Identity Information Management (EIIM) systems provide the information technology support for Master Data Management (MDM) systems. One of the most important configurations of an EIIM system is identity resolution. In an identity resolution configuration, the EIIM system accepts a batch of entity references and returns the corresponding entity identifiers. However, these batch EIIM systems lack in providing easily accessible identity information. To address this, the EIIM system is extended with an Identity Management Service (IMS) [1] to decouple identity resolution from batch EIIM processing which allows it to move into an interactive realm. Due to the uncertainty of the information provided by the requester, the system may match the reference to many different Entity Identity Structures (EIS), and the identifier returned by the system may not be the correct one. To assist the client, the IMS provides a second output called a confidence rating. The confidence rating is a measure of the likelihood that the returned identifier is accurate. The confidence rating is related to the match score, but must also consider other factors including whether the entity identity information resides in a closed or open system. This paper discusses a model for generating confidence ratings based on an assessment of the differences between the match scores for competing EIS in either a closed or open universe. The paper also includes some preliminary results from experiments performed on a production IMS system supporting student MDM. KeywordsEntity Identity Information Management; Identity Resolution; Master Data management; Open Universe; Closed Universe; Confidence rating