I have fairly new to Knime, and I was playing with the ARIMA extension as I would like to use it to predict sales. I have Microsoft’s AdventureWorkDW2012 Database as my data set. AdventureWorkDW2012 has a table called FactInternetSales which has about 60k Records; it has many columns, however, I only choose:
1- SalesAmount
2- OrderDate
Now I set the ARIMA learner’s Column Containing univariate time serious variable to SalesAmount
and everytime I try to run it, it gives the same error and you should know that I have a 32GB RAM in which 30GB is allocated for knime.
It creates matrices with similar size during training for some of the training algorithms, so 60000*60000(*8 bytes) is quite a lot (around 28GB if I am not mistaken). You can try other algorithm (after restart)/parameters, but probably it worth significant downsampling (keeping only most recent few hundred for example or selecting 1% randomly/evenly) first.
I see, and I tried it; however, it gives same value for all feature values… Anyway, can you please tell me where to find a detailed example about sales forecasting I have been searching for few days now, and I only could find two examples.
I am also getting similar error while executing ARIMA node. I had provided 1GB of Heap size. Still same issue. Is there any pre-requisite to run ARIMA extension.