I have a data set that contains records of distinct groups identified by a 'type' variable. Depending on value of this 'type' variable certain other variables are either applicable or not. Effectively these variables will be null when they not applicable...however I didn't think the usual methods to handle null values such as imputation make sense in this scenario.
One thought I had was to train separate models for each 'type' effectively getting rid of the variables that would not be applicable for each type.
I am fairly new to the world of data science and machine learning and would appreciate hearing what other methods I can consider to handle this type of data set.
