Convert the Deep-Forest model to pmml format

[zilinly]

No description provided.

[vruusmann]

Which Scikit-Learn library class that would be?

[zilinly]

from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error

from deepforest import CascadeForestRegressor

X, y = load_boston(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
model = CascadeForestRegressor(random_state=1, backend="sklearn")
model.fit(X_train, y_train)

[zilinly]

Which Scikit-Learn library class that would be?

from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error

from deepforest import CascadeForestRegressor

X, y = load_boston(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
model = CascadeForestRegressor(random_state=1, backend="sklearn")
model.fit(X_train, y_train)

[vruusmann]

from deepforest import CascadeForestRegressor

You appear to be referring to the Deep-Forest third-party package (uses Scikit-Learn API conventions, but is "organizationally" independent of it).

Given the technical complexity of the task, seems like a good candidate for extracting into an independent JPMML converter project (similar to JPMML-LightGBM, JPMML-XGBoost, etc), which could then be integrated into the main JPMML-SkLearn "umbrella" project.

Now, to help me prioritize this feature request - is there any bounty attached (eg. for getting it implemented and released in the next 2 weeks time)?

[zilinly]

Thank you so much for your response. I must say, the sklearn2pmml package has been incredibly beneficial to my work, and I'm excited to see its ongoing development. I'm particularly eager for the upcoming support for Deep-Forest. You know, it's incredibly important to many of us, and I believe it will greatly enhance the flexibility of our applications. While it may also present new challenges. I'm eagerly awaiting your future updates.

[vruusmann]

@zilinly TLDR: There is not bounty available for a quick fix, you say?

This issue is currently in "Noted" status. As things stand, it goes to my generic work queue (with no ETA).

By looking around Deep-Forest documentation I get the impression that DF models can be fully converted to PMML representation - all the necessary building blocks such as "tree models", "ensembles of tree models" and "ensembles of ensemble models" are nicely available.