Does the Match Result Include Records from the Training Set in Duplicate Detection?

·Sep 11, 2024 09:47 PM

Q10: will the "match" result also include the ones in the training set? I found out that if I put a pair of matched records in the training set and retrain the model then rerun the match again, those records that used in training set will not be marked as dup in the results.

1 comment

· Sorted by Oldest

Sonal G.
·
training sets and match results are completely different. training helps Zingg build models, which are applied on the data to give the match output results. It is likely that the models are learning/unlearning something and the results changing, But there is no logic to supress any of the training set. In fact the models built using the training set continue to be applied to different incoming datasets
✅1

Sonal G.
·
training sets and match results are completely different. training helps Zingg build models, which are applied on the data to give the match output results. It is likely that the models are learning/unlearning something and the results changing, But there is no logic to supress any of the training set. In fact the models built using the training set continue to be applied to different incoming datasets
✅1