K Nearest Neighbor Assistance

Hello Knime Community,

I need assistance with troubleshooting my algorithm. When trying to connect the Partitioning to the K Nearest Neighbor, I presented with an error message "The Dialog cannot be opened for the following reason: No column in spec compatible to “NominalValue”.

I am trying to do the following steps:

  • Partitioning Node: Using the Partitioning node to split the data into training and test sets. Set the partitioning to 70% Training and 20% Test for records 1-300.
  • k-Nearest Neighbor Node: Creating a k-Nearest Neighbor node and connect it to the Training data output from the Partitioning node.
  • Scorer Node: Connect a Scorer node to the k-Nearest Neighbor node. This will evaluate the model’s accuracy using the Test data from the partition.
  • Table Node: Attach a Table node to the second output port of the Scorer node to view the accuracy results.
  • Run Model for k Values: Configure the k-Nearest Neighbor node to run the model for k values ranging from 3 to 6.

Would someone be kind enough to check that I have not missed a step?, I have been working on this item for the past week and have search through every single source without a solution.

Thank you kindly,

Hi @abarca321

Pretty hard to help you without knowing what kind of data you are working with. The error is somewhat self-explanatory: missing the correct type of values.

I’d say have a look on the Community Hub, there are a lot of reference examples that you can use and compare to your situation :wink:

1 Like

What data type is your attribute column?

1 Like

Hi, @abarca321

Have a look at your data.
I think only numeric columns and the Euclidean distance are used in this implementation.


Thank you all for your prompt responses, the data source is coming from the following table attached.

Could you please upload some sample data? Its a lot more help than a screenshot.

Hi rfeigel,

Sure thing, please see attached sample data
BIT-445-RS-Failure-Rate.csv (7.1 KB)

Change the target to a string. That should work.

Hi rfeigel,

I am not sure i understand your response. I have multiple nodes for “string” also which target are you referring to?

Sorry. I wasn’t as clear as I could have been. By “target” I meant the column you’re trying to predict, i.e. 'failure." I cleaned up your data and changed the workflow some. This workflow produced 94% accuracy. I assumed the “model” column is a legitimate predictor. You can play with the k if you want.

Hi rfeigel,

Thank you for the example that you have shared, the last issue I am having is with the scorer, for some reason I keep getting an error with the scorer. Do you think I ran one of the nodes incorrectly? <Error Message: Execute Failed:Index 0 out of bounds for length 0>

Run the workflow I posted without making any changes. Its configured properly. I don’t know where your problem is since I can’t see your workflow.

1 Like

Hi @abarca321 .
Maybe there is a problem with your column name?


I removed it since its not a legitimate predictor. As I said - my workflow works.

1 Like

Did my workflow solve your issue?