The parameter window_size specifies a window size, but the value exceeds the

robert

Interesting error message "The parameter window_size specifies a window size, but the value 10 exceeds the example set size. "
- even though the output looks fine.

Operator #1: Read Database to load two columns, one date and one stock price.
Operator #2: I Set Role the date as id.
Operator #3: Windowing: encode_series_by_examples; horizon=1, window size=10, step size=1

As expected, I get these correct output columns, populated with the expected data:
date  close-9 close-8 .. close-0

Still the error message bothers me. It's as if the windowing thought there was no ExampleSet input at all (even window size = 1 does not solve it), but still performed the operation correctly.

Any ideas?

robert

To answer my own question:

A generated example set with identical attribute types, roles and names worked fine, i.e. I didn't get the error "The parameter window_size specifies a window size, but the value 10 exceeds the example set size.". The reason might be that the "Generate Sales Data" operator (whose output I massaged into the format I had) could supply downstream operator nodes with metadata information about how many examples it contains, whereas probably the "Read Database" operator did no such thing, as it cannot assume it will be able to select at least N examples.

Still, it would be nice if the system, at least after it ran, cleared the error message, since it gained evidence that retrieval from the database was indeed successful.

Or do you generally cache what's loaded from the DB, and how is it done most easily?

robert

To take the thread of discussion from myself, the Store and Retrieve operators did what I thought they would, and indeed the error message disappeared. So the error was clearly due to the lack of assumption that the SQL query will return any values.

tonio09

I'm also getting this error and I'm reading my data 120 000 lines with the Read CSV operator. Unfortunately, in my case I get a lot of NullPointerExceptions during execution and a bunch of errors in RapidMiner. In particular the windowing operator complains that the window size 10 exceeds the file size, and also when I do a filter data it complains that the filter index 75 000 exceeds the file size.

Later the Windowing crossvalidation throws nullpointer exception as well as the forecasting performance operator. I also have an error in RMiner about a neural net not being able to handle missing inputs.

I don't know how to resolve these errors. The file definitely has enough data lines in it. I have parallel processing enabled if that means anything.