More than one dependent variable in predicitive modelling



Hi all,

Can we have more than one dependent / target variable in predictive modelling tasks? If yes, what is the approach to be followed to predict the values for every target variable?

Thank you.


Hi @karthe1

Yes it is possible,type multimodal response, you can start by looking at glm (with Em and likelihood) or you go with method such as one vs one voting or one vs all voting.

There is even on example with xgboost in this forum, well a problem pending with the confusion matrix but it gives one example how to configure xgboost at least. xgboost does all under the hood !!

Hope this help.



You can try multivariate multiple regression for starters.
(multivariate: >1 dependent variables)
(multiple: >1 independent variables)

Another idea could be to make different models for each case if that doesn’t violate any practical requirement.


Thank you Aarshay,