Question: Suppose you have 100 training samples that you are using to train a classifier to distinguish between four classes. The training data has 50 samples
Suppose you have 100 training samples that you are using to train a classifier to distinguish between four classes. The training data has 50 samples of class 1, 25 samples of class 2, 20 samples of class 3 and 5 samples of class 4. To evaluate the stability and performance of your classifier on each class, you use 10-fold cross-validation. Is it a good strategy to randomly partition the data into 10 folds? Why or why not? If yes, fully justify why. If no, state why not, provide an alternate cross-validation scheme and justify the new scheme.
Step by Step Solution
There are 3 Steps involved in it
Get step-by-step solutions from verified subject matter experts
