LewisHotchkissDPUK
commented
1 month ago Also need to add an extra quality evaluation for relational data which is https://docs.sdv.dev/sdmetrics/metrics/metrics-glossary/cardinalityshapesimilarity / https://docs.sdv.dev/sdmetrics/metrics/metrics-glossary/cardinalityboundaryadherence
For the quality, utility and privacy functions, they should be able to handle relational table types. This means that the identifier column variable should be made optional, and in the evaluation functions the process() function from the synthetic data file should be called to identify which are the identifier columns to remove them. There should then be a loop to go through each table in the data and synthetic data dictionaries and run the standard evaluations. The results should then be combined in cases of individual scores, and averaged in case of total scores.