Comment 1 for bug 1842429

Revision history for this message
Maxime DOYEN (mdoyen) wrote :

this is right and wrong, as this detection cannot be accurate and always need a human validation, that's why I chose 'similar' txn and not 'duplicate' as a terminology.

A) a similar between 2 different files:
can be right: if there is no overlap dates
can be wrong if not, I for example have different file with some overlaps and in that case I do have real duplicates in different files

B) a similar within the same file:
if you have 3 saving account you xfer same amount at same date every month, 2 txn will be detected as potential duplicate but are in fact not

you see: if the detection occurs globally (every file into the session) or individually, some false positive can remains until a Human or IA bring knowledge and smartness into this.