Pickled Model Performs Differently in Knime & Jupyter Notebook


I used Knime to export data to an Excel sheet and trained+pickled a SVM model in Jupyter Notebook. I have loaded this model both in Jupyter Notebook and Knime and done scoring on the same data. However, the model performs quite differently even though they are using the same Anaconda Python 3 environment and scikit-learn version.

Has anyone experience similar issues? What could possible be done wrong here?


can you share jupyter code and workflow? do you mean the output probabilities/predicted values are different?

1 Like

Thanks for replying. I tried to replicate a simple case with a random table and train a classifier with it. It turns out everything is fine. Sorry for the alert.


This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.