AuotML Node Stopped Executing: Training/Testing Models failed

While developing our workflow, we encountered an issue where the AutoML Node stopped executing with an error: “Training and/or testing for all models failed, try to select more models or change input features” during the process of determining the optimal machine learning model.

Various tried options for the Model

AutoML Best Model

AutoML Best Model > Detect Models Static Predictions

AutoML Best Model > Detect Models >Breakpoint

We have decreased the number of models, but the issue is still appears.

@RishalJ welcome to the KNIME forum. Can you tell us more about your data. And would other models work and what result would they give.

And just to be sure you might try and exclude the target from the feature list.

Also try a different metric.

1 Like

Hello @mlauber71,

Thank you for your response. We are currently using this AutoML Component and plugging in this data (Attached). Our target column in “mismatch” which we are trying to predict.
processedData.xlsx (3.0 MB)

Have you tried to feed your data to a simpler model? This might help isolate the problem.


@RishalJ I just see that the size of your training set is set to 99% - maybe change that to 80 or 70 and leave out the Keras deep learning model.


Have you tried any base model before to have a benchmark?

1 Like

@RishalJ I took a look at the data you provided. You have 80 Rows, 12 Target cases and +4k features. I do not think one can build a model with this. Is this meant as a sample?


Hi @mlauber71 ,

We tried adjusting the set to around 65% to 70% along with singling out models one at a time, however it would still break either at of the Detect Models nodes. We have tried the AutoML Regression posted above and it was able to execute successfully.

@RishalJ the mismatch variable is formatted as a number so technically you can treat it as a regression problem. You should carefully examine if this is a good idea.

Also maybe convert it to a string and see if the other models would work then.

1 Like