Exporting learning model results from analysis nodes

[mlauber71]

@JaeHwanChoi pmml should be standard and be able to use in other systems. Though it seems to have some limitations

Export PMML that outputs class probabilities

I think it depends on the type of model. Normally you would have to tell the Predictor to add the class probabilities. The generic PMML predictor does not have this function - or it depends on the type of model. [image] Nodes that support Probabilities and PMML are: Additionally the Decision Tree offers the possibility to create a quite readably ruleset that you can then use with SQL or another language Additionally the Supports these model types and also has the possibility to…

A lengthy discussion about the exchange of models between knime and python can be found here:

Error when importing KNIME's Model file in Jupyter

@JaeHwanChoi the model writer will zip a given model to store it. Underneath there has to be a model that the package that you used can read(KNIME ML to Python - #2 by mlauber71 ). If you want to use it with Jupyter notebooks this has to work in both environments. You could try and store a model just as pickel and use knime as interface for some python models. Here are examples from several versions of knime and python: There are other frameworks you could use. Knime started to generic…

From my experience the best model format to interchange between knime, python, R and even Spark/BigData systems is H2O.ai MOJO:

use KNIME / Python and LightGBM to build a model - Hyperparameter tuning with...

KNIME has started to support Sklearn nodes. Though I have not tried to interchange their models with python (via pickle for example):

Sklearn: The KNIME Nodes for Scikit-Learn (sklearn) Algorithms - used to...

This seem to be the only change since we last had this discussion. The format of a model very much depends on the underlying system and packages. KNIME as a platform is trying to bring them together but I don’t think there is a universal solution or a format to cover everything. For deep learning there is ONNX but my experience here is limited.