Hands-On Machine Learning with C++

Cross-validation

We have already discussed what the validation process is. It is used to estimate the model's performance data that we haven't used for training. If we have a limited or small training dataset, randomly sampling the validation data from the original dataset leads to the following problems:

• The size of the original dataset is reduced.
• There is the probability of leaving data that's important for validation in the training part.

To solve these problems, we can use the cross-validation approach. The main idea behind it is to split the original dataset in such a way that all the data will be used for training and validation. Then, the training and validation processes are performed for all partitions, and the resulting values ...