Good question Mathieu!

One can't really use 'if-then' statements within a pipeline definition. Remember that the make_pipelinecode is really only run once to define the pipeline. This is before the pipeline has actually fetched any data. Conditional logic like this doesn't work in this context. Another way to think of it is factors such as 'eps' represent dataset objects and NOT really the actual data and therefore cannot be compared using 'if-then' logic.

So, how to construct a score based upon a number of factors with simple logic? A typical scoring mechanism is to take z-scores of each factor (to normalize the values) and then sum these up. This can be done inside pipeline with the methods and operators defined for pipeline datasets. However, a scoring system which simply adds and subtracts from a score based upon logical conditions is more problematic. Perhaps the cleanest approach is to output all the data factors and then manipulate the resulting dataframe. Generate a score in logic once the dataframe is run.

See the attached notebook for a possible solution.

If anyone has other approaches to generate composite scores it would be great hearing from you. This does come up from time to time.

Hope this helps.

Hi Dan,

Thanks so much for the detailed answer and for taking the time to reply.

I was so focused on my pipeline that I didn't even think about adding a score column to the DataFrame. Making all the calculations once the pipeline has been run make it so much easier.

I'm gonna start working that right away.

Thank you very much again!!

Cheers,
Mathieu