Open jcastro-inf opened 8 years ago
Check if a re-generated XML has differences with the original one. Check specially seeds settings.
This issue might be related with the validation (train-test of a dataset.). The replicability of the validation should be tested.
The replicability error is produced by each group formation technique. Each class needs to be evaluated individually in order to fix this bug.
When a GroupCaseStudy xml is re-executed, the replicability is ok. However, when the xml is re-generated, even if the content is the same, it produces different recommendation results. This is a serious issue, because other researchers that use delfos in the future might obtain different results.