Discussing the article: "MQL5 Wizard Techniques you should know (Part 37): Gaussian Process Regression with Linear and Matern Kernels"

MetaQuotes 2024.09.06 16:05

Check out the new article: MQL5 Wizard Techniques you should know (Part 37): Gaussian Process Regression with Linear and Matern Kernels.

Linear Kernels are the simplest matrix of its kind used in machine learning for linear regression and support vector machines. The Matérn kernel on the other hand is a more versatile version of the Radial Basis Function we looked at in an earlier article, and it is adept at mapping functions that are not as smooth as the RBF would assume. We build a custom signal class that utilizes both kernels in forecasting long and short conditions.

We continue these series on the different ways key component classes of wizard assembled Expert Advisors can be implemented by considering 2 Gaussian Process Kernels. The linear-kernel and Matérn Kernel. The former is so simple you cannot find its Wikipedia-page, however the latter has a reference page over here.

If we are to do a recap on what we covered with Gaussian Process Kernels (GPs) earlier, they are non-parametric models that are able to map complex relationships between data sets (typically in vector form) without any functional or pre-knowledge about the pair of datasets involved. This makes them ideal for handling situations where the data sets involved are non-linear or even noisy. This flexibility, in addition, makes them a bit ideal to financial time series that can be often volatile, since GPs tend to give nuanced outputs. They provide a forecast estimate plus a confidence interval. GPs help determine the similarity between two data sets and since there are multiple types of kernels to use in the Gaussian Process Regression it is always key to identify the appropriate kernel or be mindful of the shortcomings of your selected kernel particularly in instances where kernels are being used to extrapolate a forecast.