The deduplication process adds link score and driver score data to the Deduplicate transformation output. You can use the data to better understand the relationship between duplicate records.
The link score is the score between two records that identifies them as members of the same matching set. The score represents a link between a given record and the first record that it matches with a score above the threshold value. The link ID identifies the records to which a link score applies.
The link score and link ID values do not imply that a pair of records are the best match in the input data. The purpose of the link score and link ID values is to describe the composition of the matching record set.
The driver score is the score between the first record added to a matching record set and another record in the same set. The transformation uses the sequence ID or row ID values to identify the first record in the set. Driver scores provide a means to assess all records in the set against a single record.
Note: Duplicate analysis generates a single set of scores for the input records. The driver scores and link scores represent the different relationships between the records and do not indicate different types of duplicate analysis. The driver score and link score assignments can depend on the order in which the records enter the transformation. A driver score for a given pair of records might be lower than the threshold value.
Example of link scores and driver scores
A Deduplicate transformation analyzes records with a column of surname data. The deduplicate asset defines a threshold value of 0.825 for duplicate records.
The following table shows the results that the transformation might return:
Surname
Sequence ID
ClusterId
ClusterSize
DriverId
DriverScore
LinkId
LinkScore
SMITH
1
1
2
1 - 6
1
1 - 1
1
SMYTH
2
2
2
1 - 3
0.83333
1 - 2
1
SMYTHE
3
2
2
1 - 3
1
1 - 2
0.83333
SMITT
4
3
1
1 - 4
1
1 - 4
1
SMITS
5
4
1
1 - 5
1
1 - 5
1
SMITH
6
1
2
1 - 6
1
1 - 1
1
The results provide the following information about the surname data:
•SMITT and SMITS do not match any other record with a score that meets the threshold. The transformation determines that the records are unique in the data set. The transformation can assign score values of 1 to the records because each record matches itself uniquely.
SMITT and SMITS each have a ClusterSize value of 1, which indicates that they are the only record in their respective sets. To find unique records in the output, search for matching record sets that contain a single record.
•SMITH and SMITH have a link score of 1. The transformation determines that the records are identical. The transformation adds the records to a single matching record set. The ClusterId value indicates that the records belong to the same set.
•SMYTH and SMYTHE match with a score of 0.83333. The score exceeds the duplicate threshold. Therefore, the transformation adds the records to a single matching record set.