Multi output random forest regression

psagrera

Hi All,

is it possible to create a multi output random forest regression using a visual analysis ? AFAIK you only can select one feature at a time to be predicted.

I don't know if subpopulation analysis is the way to go here or it must be done via code recipe.

Thanks

psagrera

I will reformulate the question .

Having a dataset with the following shape:

- Training X_train : (33873, 17) y y_train :(33873, 15)
- Validation_1 X_test : (14517, 17) y y_test : (14517, 15)
- Validation_2 X_test2 : (145973, 17) y y_test2 : (145973, 15)

So, let's going to predict 15 values using random forest regressor. is it possible to do that using visual analysis ?

Thanks

Darien

Hello,

You are correct, it is not possible to build a multi output model using a visual analysis. You can however create a custom model.

To do this in DSS:

https://academy.dataiku.com/custom-models-in-visual-ml

Documentation:

https://doc.dataiku.com/dss/latest/machine-learning/algorithms/in-memory-python.html#custom-models

Multi Output RF Model (may be helpful?):

https://scikit-learn.org/stable/auto_examples/ensemble/plot_random_forest_regression_multioutput.html

N

@DarienM
What datatype should the target be in this case? I don't see how to do this with custom model becasue the target output has to be numeric. It does not support numpy.ndarray or python list.