Using Rapid Miner Go, applying model to other data sets, it is not recognizing dates - says Missing

jsdrewjsdrew Member Posts: 9 Learner I
I've used Rapid Miner Go to create models for my data.  When I apply the model to other data sets there are three columns with dates that come back as "MISSING" in the prediction results, but were clearly there in the dataset.  Any idea what is going on here?

Best Answer

Answers

  • KristofGasparKristofGaspar Employee, Member Posts: 3 RM Engineering
    Hi, 

    Can you please provide a screenshot of your browser view with the id visible in the url to speed up investigation?

    Attached an example.
    Thanks.
  • jsdrewjsdrew Member Posts: 9 Learner I
    I've attached two files.  DataSetforGradient.PNG shows the dataset after it has been uploaded into RapidMiner Go and prepared for the model to be applied to it. You can see the two fields labeled "Latest..." have date data in them.  PredictionResultsforGradient shows the results after "Calculate Predictions" is clicked on the screen shown in DataSetforGradient. For some reason the Prediction Results are labeling most of the data in the "Latest..." fields as "MISSING".  

    Thanks for your help.
    Sam Drew
  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,254 RM Data Scientist
    Hi @jsdrew ,
    it seems the dates in your dataset where wrongly identified as a string / categorical value, instead a date.That causes a misbehavior, since auto model does not use the proper preprocessing for dates. What happens is, that your model only works on dates which where present in the training set, but is misbehaving on dates which were not.

    Best,
    Martin
    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
  • jsdrewjsdrew Member Posts: 9 Learner I
    So it is dates in the training set that are the problem?
  • jsdrewjsdrew Member Posts: 9 Learner I
    I've gone back and looked to verify that the dates in the new dataset were present in the training set, and I am still having issues where some categorical data comes back MISSING. 
Sign In or Register to comment.