Hands-On Artificial Intelligence for Beginners

K-fold cross-validation

You've already seen a form of cross-validation before; holding out a portion of our data is the simplest form of cross- validation that we can have. While this is generally a good practice, it can sometimes leave important features out of the training set that can create poor performance when it comes time to test. To remedy this, we can take standard cross validation a step further with a technique called k-fold cross validation. 

In k-fold cross validation, our dataset is evenly divided in k event parts, chosen by the user. As a rule of thumb, generally you should stick to k = 5 or k = 10 for best performance. The model is then trained and tested k times over. During each training episode, one k segment of the data ...