Backward Feature Elimination
In backward feature elimination, we start with whole set of features and we remove the one with least amount of importance.
The flow chart of backward elimination is given below:
On another variation, on each iteration we remove each feature and train the model to find out the importance of that feature in the accuracy. We remove the one which has a least amount of affect on the accuracy. But it is computationally expensive as we have to train the model multiple times on each iteration which is infeasible for a large dataset or complex model.