ARIMA Learner Node Inquiry

Hello KNIME Team.

I have a question about nodes related to time series analysis.

Among the time series analysis nodes provided by KNIME, ARIMA Learner is very tricky to use.

When I run the training with about 10,000 data, the error message “Java heap space” appears, and when I change the estimation method to mle, the error message “Not enough memory to learn with Maximum Likelihood method” appears.

I finally understood that it means there is not enough memory, so I maximized Xmx in knime.ini and got the same error.

Is it impossible for users to utilize ARIMA Learner with more than 10,000 data on KNIME AP?

Thank you.

Hi @JaeHwanChoi,

I’d like to start by saying I wouldn’t recommend using an ARIMA model on 10,000 rows of data - it typically functions a bit better on smaller data sets.
Depending on your specific use case you could either use a smaller subset of the data for training, i.e. the final 1,000 rows. Or you could aggregate your data to a lower granularity.

If beyond the old ARIMA node we also have a set of Time Series Analysis components that you can try. You can read a bit about the SARIMA one here:

2 Likes

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.