A new Just KNIME It! challenge just came out on explainable AI (xAI). Let’s explore the LIME technique with this data puzzle to make research on breast cancer prediction models more transparent.
Here is the challenge. Let’s use this thread to post our solutions to it, which should be uploaded to your public KNIME Hub spaces with tag JKISeason3-18 .
Need help with tags? To add tag JKISeason3-18 to your workflow, go to the description panel in KNIME Analytics Platform, click the pencil to edit it, and you will see the option for adding tags right there. Let us know if you have any problems!
Here’s my solution. Rather than trying to completely reinvent the wheel, I used the LIME workflow from the Hub. It required a lot of revision, but seems to work fine. Probably could be improved with some parameter optimization.
Hi all,
Here is my solution. I used a simple random forest without any optimizations. I calculated the average LIME value for each class, as well as the LIME values for each sample, to determine which features might contribute to the predictions. I believe that optimizing the prediction conditions could alter these results. Thank you.
Hi all,
Here is my solution. I created a generic LIME analysis workflow for the binary classification model. This workflow will work for different datasets as long as the appropriate target column is selected in the Column Settings component.
In this workflow, you can focus on False Positives and False Negatives to analyze which features contributed to the incorrect predictions. This will be useful for improving the model and understanding the quality of the data.
Getting one on the board as well… with most machine learning topics still a bit out of my depth, but amazed how one can pull things off by following examples.
In order not to just copy any other solution I set up parameter optimization for Gradient Boosted Tree and then wrangled the data so that I could use some of the visuals from the LIME Example on the hub (Bubble and Violine plot), plus some of the bar charts that were used here.
A new Just KNIME It! challenge just came out on explainable AI (xAI). 
Let’s explore the LIME technique with this data puzzle to make research on breast cancer prediction models more transparent. 
Need help with tags? To add tag JKISeason3-18 to your workflow, go to the description panel in KNIME Analytics Platform, click the pencil to edit it, and you will see the option for adding tags right there. 
Let us know if you have any problems!
Another Tuesday, another solution to our beloved Just KNIME It! challenges! 
Once again you folks went above and beyond with the modeling and visualization for this challenge. Bravo!! 
Here’s our solution highlighting the power of LIME.