none
Capstone- Data Science track (Predicting gross rent) RRS feed

  • Question

  • Hi,

     

    I am still stuck trying to score my test prediction labels. Sometimes I get the "scored labels" in the test prediction. But the final evaluation fails indicating there are no label columns in the scored test data, which is true because I don't expect the test dataset to have the label columns, in this case, gross_rent. But I do expect the predicted scores to have those labels. When I decide to include an empty column of gross_rent into the test dataset, it rejects it and accuses me of having empty columns. In some cases, I did not get any scores at all, like this error message:

    "AFx Library library exception: table: The data set being scored must contain all features used during training, missing feature(s): 'gross_rent'.  (Error 1000)".

    Why does it expect me to include a label column in the test dataset when that is what I am trying to predict with the test dataset (using the trained model)? And even when I obey and add that feature it again tells me I have empty label columns. I have cleaned up my data in trying to make sure I satisfy all doubts. Is there something I might still be missing? I have been stuck for 5 days with the same issue and cannot submit my report (due tomorrow!).

     

    My trained model hovers between 0.79- 0.85. The different numbers are due to trying everyway to beat that scoring error.

    Thanks in advance for your help.

    Monday, December 30, 2019 8:48 PM

All replies

  • Hello,

    Is it possible to publish your experiment to the gallery to check if this can be replicated?

    It looks like your scenario might be similar to the one mentioned in this thread

    Basically, if you are using the predictive experiment with an input data every call to the prediction web service might be re-creating the input based on the import data source. 

    -Rohit


    Tuesday, December 31, 2019 11:31 AM
    Moderator