0
$\begingroup$

I have a dataset where some features (c,d) apply to only when a feature (a) is a specific value.

For example

a, b, c, d
T, 60, 0x018, 3252002711
U, 167, ,
U, 67, ,
T, 66, 0x018, 15556

So I'm planning to splitting the dataset so that there are no missing values.

a, b, c, d
T, 60, 0x018, 3252002711
T, 66, 0x018, 15556
a, b
U, 167
U, 67

and then put these into individual models which combine to produce a prediction. Is there a way to do this? Because I've done some research on ensemble learning and it appears to only be for datasets with the same set of features.

$\endgroup$

1 Answer 1

0
$\begingroup$

This is not ensemble learning, it's just subsetting the dataset based on different cases, like a piecewise function.

Basically you obtain two datasets based on your condition, you train an independent model for each of them. Then when you want to predict for some new test instance, you select the model to apply based on the same condition. If needed the test predictions from the two models can be joined back together for evaluation.

Of course it's important to have a sufficiently representative sample as training data for each model.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.