Hi All,
@AlexanderFillbrunn, @mlauber71
I have done the XGBoost modes now i want to save that in PMML format, but its not allowing me to do so. Please let me know how i can do the same or is there any other way to get this done.
There is only a limited number of models that can be converted to PMML. XGBoost is not among them as of now.
Hi,
@mlauber71
Is there any way, i can get this rules like Decision Tree for XGBoost so that i can put that in production or how can i put this into production?
Yes you can by saving the model with the model writer producing a proprietary zip file and using that.
XGBoost with H2O.ai is not in KNINE yet? The you could put that to Spark also.
I would have to check if the MOJO predictor could use XGBoost models produced by h2o automl.
Hi,
if you use the Gradient Boosted Trees Learner, you may also be able to utilize the Gradient Boosted Trees to PMML node to generate PMML.
Kind regards
Alexander
1 Like
Thank you very much.
Currently i am facing issue while reading the hyper parameters in flow variable tab (Not sure the meaning of those), hence struggling to input their values in Gradient Boosting.
Would you please let me know some document/s that lists out all hyper parameters and their meaning for Gradient Boosting as per Knime?
Hi @ChetanP,
unfortunately there is no such document. Which hyperparameters do you want to control specifically? I could find out for you which flow variable mappings to set.
Kind regards
Alexander
Hi,
That would be great.
I am sharing the excel sheet. Please mention Knime analogous name in column “c”.
Thanks in advance!!GBM_Parameters - Knime.xlsx (12.3 KB)
Hi,
here are the names for flow variable configuration. Not all options are applicable for our node. I have indicated that in the table.
Kind regards
Alexander
GBM_Parameters - Knime.xlsx (12.4 KB)
1 Like
Hi,
Would you please check and let me know please?
Thank you very much. I would appreciate your help if you could do the same for XGBoost (i have done it though just to be sure).
I am attaching excel sheet for XGBoost hyper parameters, i have attached Knime Analogous name for all hyper parameters (in column c)XGBoost Parameters - Knime.xlsx (17.8 KB)
,just to be sure, please cross check once
Hi,
we don’t have a specific XGBoost node. For that you need to use the Python Snippet node.
Kind regards
Alexander
I am trying to tune the hyper parameters of Gradient Boosting. Would you please let me know what is significance of michildsize? Is this to control number of records that enters in the terminal node? Because i want to tune that as well
Hi,
even though this flow variable does not correspond to a setting in the dialog, it seems to work. So minChildSize is, as you correctly assumed, the number of nodes in a terminal node.
Kind regards
Alexander
I checked and at the moment it is not possible to load H2O.ai MOJO files generated via their Python (or R) packages into KNIME and reuse them. Although it is possible to use a Python node to create scores from them. But that may not help if you want to distribute them via KNIME to a server or Big Data cluster like with the other MOJO models.
Thank you very much
@AlexanderFillbrunn , @mlauber71
I am trying to tune the parameters using Bayesian Optimization but not sure importance/tune the parameters in it. Please let me know how can i use (one line explanation will help me here)
Random seed, Enable step size, Max. number of iterations, number of warm-up rounds, Gamma, Number of candidates per round.
Hi,
Have you checked the node description? There should be an explanation for every option available.
Kind regards
Alexander
Hello,
No i am not sure. Would you please let me know how can i check the node description & where is it available?
Hi,
You finde the node description either directly in the Analytics Platform (View -> Node Description) or on KNIME Hub:
Kind regards
Alexander
1 Like