Solutions to "Just KNIME It!" Challenge 26

This thread is for posting solutions to “Just KNIME It!” Challenge 26, wrapping up our 4-week series on data classification! How can we communicate our model’s performance visually? :bar_chart: :chart_with_upwards_trend:

Here is the challenge: Just KNIME It! | KNIME

Feel free to link your solution from KNIME Hub as well!

And as always, if you have an idea for a challenge we’d love to hear it! :heart_eyes: Tell us all about it here.

This is my attempt for challenge 26

I’m just exploring the data visualization nodes as well as the Local Explanation View component and how to implement it in the workflow.

Data Visualization

Local Explanation View Dashboard

1 Like

I leave it up to others to elaborate more on their solutions and provide more details on how to interpret the results of the Local Explanation View and other visualization dashboards

1 Like

Hey @si_daniel_a, thanks for your contribution! You may find our blog post on local explanations useful, by the way. :slight_smile:

This is my solution for the Just KNIME It challenge 26

The screenshots show the results of the local explanation and Global explanations

1 Like

Hi everyone,
Here is my solution.
ROC curves for Churn (P=0, P=1) and scorer output.

1 Like

Here’s my solution. The rule engine filters false pos/neg predictions to feed the local analysis.

1 Like

My Submission for Challenge -26


@AnilKS I like your solution, but I see that you trained a Gradient Boosted Model and scored it and explained its accuracy and other metrics with bar chart, but while explaining the global and local feature importance you trained an AutoML.

I wanted to showcase how you can explain your already trained Gradient Boosted model using the Global and Local Explanations Component in KNIME. I used your solution and removed the AutoML component from it and just added Integrated Deployment - Capture nodes to explain your existing model.

Let me know your thoughts on this, the workflow can be found here

1 Like

:sunny: Hello, everybody! :sunny:

As always on Tuesdays, here’s our solution to last week’s #justknimeit challenge. We created a component to answer question 1, allowing for an interactive view that lets you compare different performance metrics for both classes :bar_chart: Next, we used the Global Feature Importance verified component to answer question 2, and created another component to give local explanations to both a false negative and a false positive test instance. :mag: This last component is heavily based on the Local Explanation View verified component.

I hope you learned a lot about model inspection while creating these visualizations! :wink:

:boom: See you tomorrow for a new challenge! :boom:

Dear Mahantesh ,
Really glad and thankful to see approach… i chose to use the simple graph as the current output nodes weren’t compatible for the feature importance nodes… Appreciate your help and support to suggest the extended approach ( with workflow executor nodes ) for existing problem of single models .Kudos to knime team . Who are there all the time.


Thank you for the kind words, @AnilKS! We try… :slight_smile:

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