default for train/validation split in fit_class_random_forest

[soxofaan]

openeo-processes/proposals/fit_class_random_forest.json

Lines 27 to 33 in e4df864

	"name": "training",
	"description": "The amount of training data to be used in the classification, given as a fraction. The sampling will be chosen randomly through the data object. The remaining data will be used as test data for the validation.",
	"schema": {
	"type": "number",
	"exclusiveMinimum": 0,
	"maximum": 1
	}

Maybe this has been discussed before, but can't we pick a sensible default for the training-validation split? e.g. 80% / 20%?

[m-mohr]

Sure, what is a sensible default?

[soxofaan]

Python sklearn's train_test_split seems to default to a split of 75% train - 25% test/validation

But most articles I checked from a quick google suggest 80% train - 20% validation as good default.

[m-mohr]

dev telco: Remove this parameter completely