Enforce FindIncrementalMatches in AWS Glue to use existing match_id
The use case is that we have an incremental source of data from which we need to identify the matching records. For this purpose, we are running Find Matches using AWS Glue 2.0.
Result after Find Matches is Run
When I run FindMatches on the initial source, the following result is generated against the source. Note the match_id generated for each record
Result after Find Incremental Matches is Run
When I run FindIncrementalMatches using the above result of Find Matches as the existing source, the following result is generated with completely different Match IDs:
Question
Is there a way to enforce FindIncrementalMatches in AWS Glue to use existing match_id while processing matches from an incremental source?
Important links
We've followed the steps described in the following link:
- 최신
- 최다 투표
- 가장 많은 댓글
I understood that you wanted Glue to use the same match_id's for subsequent FindIncrementalMatches runs for incremental data.
Please note that the match_id's are arbitrary identifier which act as labeling for your data and it denotes the matching records which is predicted by your ML transform Algorithm. In case of incremental data the datasets is changing overtime and new or or modified records are being added thus there are now more candidate rows taken to be consideration by the machine learning model to decide the pairing and it is possible to have new match_id as it starts labeling from scratch, just like you provided as in the example. Unfortunately in Glue we do not have any option to enforce the previously generated match_id's to put it in use for the subsequent FindIncrementalMatches runs for incremental data as of now.
Thank you for the response. I have a follow-up question that there is a field with the key: "enforcedMatches" in FindIncrementalMatches. What is the purpose of this field?
It is mentioned in "https://docs.aws.amazon.com/glue/latest/dg/glue-etl-scala-apis-glue-ml-findincrementalmatches.html" but there is no implementation example for this.