How the variable importance is measured during model building?

r
random_forest

#1

Whenever I have built a model using different algorithms like the random forest I usually calculate the variable importance.So I do not overfit the model.

variable importance - It tells about the predictor which contribute most in predicting the dependent variable.

There is function which we can use to calculate the importance of the variable like in random forest importance()

I want what is the logic for this method .


#2

Hi @ankit81195, could you elaborate on [quote=“ankit81195, post:1, topic:8424”]
I usually calculate the variable importance.So I do not overfit the model
[/quote]


#3

It is good to have the explanation for the one which @jalFaizy mentioned.

On the variable importance side, it is computed based on how much the split on the given variable improves the model performance such as reducing the squared error or node impurity and so on.

A detailed explanation is present in Chapter 10.13 of this book, Elements of Statistical Learning

Hope this helps.

Thanks,
SRK