[question] Chapter-2 (page-no:91)-Random Forest Regressor training of the dataset(housing)

[prabhjot26-jangli]

Enter the chapter number

chapter 2 End to end MAchine Learning Project

Enter the page number

Page 91 , RandomForest Regressor

What is the cell's number in the notebook

in the line , [ if you train a Rando Forest and measure the RSME.........],

Enter the environment you are using to run the notebook

Jupyter on Windows

Question

My question is : The author says let's try one model now, the Random Forest regressor. And he uses it. And then performs a cross validation using the variable forest_rmses. After the test, he again says: [ However, if you train a random forest and measure the RMSC on the training set, you will find roughly 17474]. So now the question is. Are these both different because in the previous forest_rmses score...the RMSE is 47019

[ageron]

Great question, thanks. The first evaluation uses cross_val_score() which performs cross-validation across cv splits (in this example, cv = 10). This means that the model is trained on 9 splits and evaluated on the 10th, so it's a fair evaluation, we're evaluating the model on data it hasn't seen during training (and we do this 10 times, each time changing the evaluation split). But when we evaluate the model on the training set, we see that the error is much lower, which means that the model is overfitting the training set quite severely. I hope this helps.