"No useable observations" error with HP Neural in EM

bmutell — Thu, 30 Mar 2017 18:31:24 GMT

I'm running an HP Neural node in Enterprise Miner to predict toward a multicategorical nomial target. I receive the following error message due to a run time error:

NOTE: There were 57066 observations read from the data set EMWS1.PART_TRAIN.
NOTE: There were 24457 observations read from the data set EMWS1.PART_VALIDATE.
NOTE: Reading data...
ERROR: No useable observations.
NOTE: The SAS System stopped processing this step because of errors.
NOTE: There were 81523 observations read from the data set WORK.HPNNA_TRAINDATA.
WARNING: The data set EMWS1.HPNNA_OUTMODEL may be incomplete. When this step was stopped there were 0 observations and 0 variables.
WARNING: Data set EMWS1.HPNNA_OUTMODEL was not replaced because this step was stopped.

I have been unable to find any documentation online as to the cause of this error message. I have run the same data through several decision trees and random forests with no hassle.

Thanks in advance.

Re: "No useable observations" error with HP Neural in EM

WendyCzika — Thu, 30 Mar 2017 19:26:30 GMT

My best guess is that each observation has at least one missing value for an input or target. Can you see if there is an input with a lot of missing values and either reject that variable, or use HP Impute to impute?

topic Re: "No useable observations" error with HP Neural in EM in SAS Data Science

"No useable observations" error with HP Neural in EM

Re: "No useable observations" error with HP Neural in EM