How many columns in your testSet? Does it has y or not?
And what should the pred of testSet looks like?
I may need this size to design the dataEngineer function.
Your predict function outputs a list of predicted values, one per input row. This is standard and there is an example in the tutorial notebook.
You can of course change the cardinality of the data you directly feed into your predict function (I have called it K on the following example). However, we should be able to do this:
Take the raw dataset minus the target (N x 164), pass it in your dataEngineer function, take the output from dataEngineer (N x K) and call predict on this output (N x K), so that it gives exactly 1 prediction per row, i.e. a N-long list.
Thanks Herve,
pred = myPredictionFunc(testSet, engineerData, model)
So your testSet here is Nx165?
we have the true labels indeed, but your model should obviously use 164 features to predict the target value.