No video
Should I shuffle samples with cross-validation?
Рет қаралды 844
Күн бұрынBy default, the cross_val_score function in scikit-learn does not shuffle samples. In this lesson, you’ll learn when you might need to shuffle and how to do it.
P.S. This is a lesson from my NEW course, "Master Machine Learning with scikit-learn." You can enroll here: courses.datasc...
For all paid courses, I offer location-based discounts (up to 75%) to people in 160+ countries. Check your discount here: courses.datasc...
Enroll in a FREE Data Science course here: courses.datasc...
@dataschool 3 ай бұрын
This is a lesson from my NEW course, "Master Machine Learning with scikit-learn." You can enroll here: courses.dataschool.io/master-machine-learning-with-scikit-learn
@sedighehnadaei1895 2 ай бұрын
As always you did great.thank you so much ❤
@dataschool 2 ай бұрын
You are so welcome!
@aleksandartta 3 ай бұрын
Hello Kevin, thank you very much... I have two questions: 1) after hyper parameters tunning and cross validation, the final model should be some that is trained on the whole dataset (meaning train + validation set)? Am I right? 2) do we need cross validation if the dataset is very big (and how to know how big :) ? i.e. when cross validation is not necessary?
@dataschool 3 ай бұрын
Great questions! 1. Yes, re-train the tuned model on the entire dataset (meaning all samples for which you know the target value). 2. Yes, cross-validation is a useful model evaluation procedure with any size dataset, with the possible exception of a very tiny dataset. (Below a certain number of samples, no model evaluation procedure is particularly useful.) Hope that helps!