I'm looking to employ a blended approach using, potentially, least squares linear regression for a classification problem. The level-1 classifier should output a probability between two classes. Could I train the level-1 when the level-0 classifiers are trained using two distinct data sets? From my reading, the usual approach is training all classifiers with the same data set.
Background: One is collecting motion sensors and the other pressure from a touch screen. They both output a percentage.
If the answer to the question above is no, how could it work or what would be a better approach?
