Hey folks,
Wanted to share with you a demo on how to install and run KNIME inside Kaggle kernel (or ANY other linux environment without GUI). It covers:
- downloading and unpacking the runtime;
- changing memory settings for the runtime (from command line);
- installing KNIME extensions from command line (Python and H2O integrations)
- deploying workflow (developed with KNIME GUI);
- installing and using the wrapper python package (or KNIME command line processor directly)
Apart from that - the workflow demonstrates good FE practices and utilization of advanced ML algorithms (like LightGBM), model ensemble etc. It is amongst the very well scoring public kernels on this playground competition (Categorical Feature Encoding Challenge II).
This is also to demonstrate, that it is possible (and feasible) to use KNIME even in a code only Kaggle competitions (and Kaggle competitions in general - as I do with pretty good success for years).
Here is the kernel itself:
https://www.kaggle.com/gpamoukoff/using-knime-in-kernel-complete-pipeline-glm-lgb
Hopefully you’ll liked it - and you saw something new today. Let me know if you have any questions in the comments. Cheers 