Creation of NN Model and Logistic Regression Model

Hi All, I am trying to run an NN Model and Logistic Regression Model as per the attached Synthetic Data and the exported models created by myself. Though, I am not getting the right accuracy.

Therefore, need some support/guidance to understand if I am doing it correctly or if I am somewhere doing it incorrectly.

Pooja Sharma
NN Project_Questions.docx (677.2 KB)
HR_DATA.knwf (60.2 KB)

@knimediger @ActionAndi Would you be able to advice on this query. Thanks

Hi @psharma_18 -

You haven’t included your dataset, so we can’t execute your workflow. Either upload or the dataset itself, or re-upload the workflow in a non-reset state.

Although I will say that since this is clearly a class assignment, you probably aren’t going to get a direct answer from our experts, but instead maybe some hints that will help you. :slight_smile:

When you say you are “not getting the right accuracy”, can you be more specific?

1 Like

Hi Scott,

Yes, its not an active class assignment. I am myself doing it online to understand if my understanding is going well or not. I have created the workflow, though seems I am missing the understanding part on the Independent Variables to be considered in a workflow.
Attached is the required Data Set.

Data.pptx (49.9 KB)

A PowerPoint with a binary file included is not a good format. Such files might contain unwanted code. You should probably just ZIP the File or include the data with the workflow so one can see the settings.

If you are interested in learning more about machine learning I have these links to offer

And som additional links from a collection - mostly from the KNIME forum.

HRDataset.xls (153.5 KB)
Please check if this format works. I have gone through the sessions and on the basis of that I have created these 2 workflows. Though not confident if I am doing it correctly.

What is the “right accuracy”? What percentage do you currently get?

With NN Model, I am getting the accuracy of 75.806% and with Logistic Regression its showing 66.129%.

And what is the benchmark? Is it possible to get a higher accuracy with the data? Have you tried ensembles? Did you do any hyperparameter tuning?

Hi Daniel, I am not aware of any benchmark. I did the tuning of the hyperparameter. Though I haven’t used the ensembles.
Also, as this is a Synthetic data that’s why not sure if this is enough or not. However, as per my understanding with Correlation, I think there should be more data required.

Sorry, but I fear I’m not the right contact to support in this topic. Maybe you wanted to add @knime ?

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