Data Science

What is cross-fold validation?

Instead of splitting the data set in just holdout data and training data, cross-fold validation splits the set in different partitions (folds) and iteratively chooses one of the partitions as holdout and the other one as testing set.
 
This balances out the accuracy across different partitions and avoids results, that are just due to chance.
 
- Each fold is used once to generate a model and k-1 times for testing a model

Diskussion