This video describes ways of combining outcome of multiple ML models to improve predictive performance through Voting, Averaging and Stacking. For more such content, subscribe to our newsletter on machinelearnin...
As you said model m1 to m4 work better for some set of data. Input features are also given to meta model and meta model will decide 'ok for these features m2 worked well so lets give more weightage to m2 output'. it raises question that how meta model will classify that for which input features which model performed well?? Hope you got the question