05-26-2016 01:20 PM
How does EM handle categorical predictor variables for all modeling nodes? Does it do encoding transforming like Linear Regression? If this is the case, will encoding affect prediction power for Neural Network, Decision Trees, Gradient Boosting etc.?
05-31-2016 11:56 AM
Neural networks do dummy encoding of categorical/class inputs like a regression model.
Tree-based algorithms try grouping levels in every possible combination when determining the best split for each node of the tree.