Machine Learning: Gaussian Process Regression

Machine Learning: Gaussian Process Regression distills recent price data into a smooth fitted curve and extends it into the future, delivering trend estimation and forecasting in a single plot. It is a Bayesian relative of kernel smoothing, which is why the output reads as a clean line through noise, and why it anchors the library's machine-learning lineup.

How to Trade the Machine Learning: Gaussian Process Regression?

• Fitted line: condenses the Training Window into a clear local trend, ideal for descriptive analysis and pattern study free of noise.
• Forecast extension: projects that trend a set number of bars forward; read it as one plausible path forward, not a promise.
• Lock Forecast (default): the projection stays static, so you can grade it against what actually happened.
• Update Once Reached: the forecast refreshes when its window completes, staying relevant as price evolves.
• Continuously Update: recomputes on every bar for traders who want the current read at all times.

Smooth stretches the estimate toward longer-term trends, while lower Sigma values amplify the forecast, with the caveat that amplification can intensify errors, particularly on larger Training Window values.

Machine Learning: Gaussian Process Regression Settings

• Training Window: the scope of recent price data the model fits.
• Forecasting Length: the future bar count for the projection.
• Smooth: the model's smoothness degree; higher values yield more extended, less volatile estimates.
• Sigma: noise variance controlling sensitivity to outliers and forecast amplitude.
• Update Mechanism: chooses between Lock Forecast, Update Once Reached, and Continuously Update.

Frequently Asked Questions

Can I control which bars the model trains on?

Are the forecasts reliable enough to act on alone?

What does it take to access this indicator?