Hands-On Machine Learning for Algorithmic Trading

The bias-variance trade-off

The errors that an ML model makes when predicting outcomes for new input data can be broken down into reducible and irreducible parts. The irreducible part is due to random variation (noise) in the data that is not measured, such as relevant but missing variables or natural variation. The reducible part of the generalization error, in turn, can be broken down into bias and variance. Both are due to differences between the true functional relationship and the assumptions made by the machine learning algorithm, as detailed in the following list:

• Error due to bias: The hypothesis is too simple to capture the complexity of the true functional relationship. As a result, whenever the model attempts to learn the true ...