Compute missing values on a training data set and impute them on a new data set. Current available options are median/mode and random forest.
