Virtual experiments in input range detection: how to proceed

Dear KNIME community,

I am working on a Data Analytics project and thanks to the KNIME team I managed to generate virtual data so I could train the models better. The workflow they showed me was the following from the KNIME Forum: Data Generation Example: Supermodels)

Using the workflow they sent me the Decision Tree worked well as a preliminary approach, predicting good/bad parts (using 3 input variables).

The aim of the analysis is to identify the ranges of the inputs which will lead to good parts. A Design of Experiment have been conducted and the input variables have the following shape:

The output variable is a OK/NO OK type, meaning good/bad part.

I want to identify the combination of ranges of the input variables which will lead to good parts, so I can understand better the manufacturing process.

I hope you can guide me in the appropriate workflow or which nodes are the ones I need to use. I heard that a possibility might be virtual experiments but I am not sure how to apply it using KNIME.

Thank you very much in advance.
Kind regards,

Hi @Aner and welcome to the forum.

This sounds like a good use case for model interpretability and/or explainable AI - basically, trying to understand which variables have the most impact on given predictions.

Iā€™m going to drown you in links for a minute. Apologies in advance :slight_smile: The good thing is that each of these blog posts has example workflows and components that you can try out on your own.

1 Like

Thank you very much for your response @ScottF,

I will check the posts and try the workflows and nodes on the dataset.

Thanks again,

1 Like

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