What the best practice to reuse Knime model in Python?

[howellyu]

Hi there,

I’m trying to load a Random Forest Model trained in KNIME into Python.

I have used 2 ways to output the model:

1. 
2. 

However, I am having trouble loading the trained model into Python.

Method I have tried:

1. Python Pickle loader.
2. Python sklearn_pmml_model package

Python Version: 3.6
Knime Version: 3.7.0

Any response will be appreciated. Thank you!

-H

[mlauber71]

Here is an example with Python an pickle. For your Szenario question would be to generate the model without python first.

Will Knime support saving or loading the model in pickle format?? I think the python script only support jython code as per documentation can someone tell me whether it will also support python code.

Yes I can. The attached workflow uses Python Random Forest Learner, stores the model as Pickle file and reads it again. The structure with Object Writer is mainly there to follow a common KNIME structure. Technically it also does work to just do everything within a Python Script node. Not sure about jython in that context. Pickle could be written as binary (“wb” write binary). [image] kn_example_python_census_income.knar (1.4 MB)

Best practise I came across was using H2O.ai MOJO files.

H2O Model to MOJO — NodePit

They mostly can be exchanged between KNIME and R and Python. I would have to set up an example.

A general example can be found here:

Any advice to improve the performance of a classification model ?

We had a similar discussion about metrics in multi class predictions here: Besides accuracy you could use a LogLoss metric. That means that not only the outcome of prediction will be taken into account but also the ‘level of certainty’. If you have two predictions from two models A) 1=51%, 2=39%, 3=10% TRUTH=>1 B) 1=90%, 2=5%, 3=4% TRUTH=>1 both come up with the correct result (1). But prediction B) would be better as it predicts the correct answer with higher confidence and might be mor…