Cross-Validation
From Open Risk Manual
Contents
Definition
Cross-Validation is an approach to empirical Model Validation of statistically based risk models using historical data.
Validation Set Approach
The simplest validation set approach involves dividing the available data into a training set or Development Sample and a validation set or Hold-Out Sample
Leave-One-Out Cross-Validation
This approach involves splitting the data set repeatedly into two parts: Instead of creating two subsets of roughly comparable size, a single data point is used for validation and the remaining data make up the training set.
k-Fold Cross-Validation
Similar to the Leave-One-Out approach but instead of a single data point, the sample is split into k roughly equally sized folds and in each iteration one of them is used for validation