NaN issue when multi-target regression

[atwahsz]

the autoxgb study does not launch when one of the targets is missing .

is there anywork around ?

[abhishekkrthakur]

Why would anyone have missing targets in training data? :)

[atwahsz]

not completely missing . some rows will have 2 targets and one target is Nan . ideally that row is eliminated for the third target. if a single target model is built

[abhishekkrthakur]

The target should never be missing. If the target is missing, you need to handle them before inputting to autoxgb. Not just xgboost, any model will fail if targets are missing :)

[INF800]

Are we using any kind of transformations on target variables? For example if we have skewed log normal distribution basic convention would be to apply log transformation

[abhishekkrthakur]

no. There is no transformation of target variables.