Skip to main content
In time series forecasting, it is important to consider the full probability distribution of the predictions rather than a single point estimate. This provides a more accurate representation of the uncertainty around the forecasts and allows better decision-making. TimeGPT supports uncertainty quantification through quantile forecasts, prediction intervals, and forecast simulation. Quantiles and intervals describe the uncertainty at each future timestamp; simulation goes one step further and generates complete future paths, so you can answer questions about outcomes that span the whole horizon, such as total cost or the chance of ever crossing a threshold.

​
Why Consider the Full Probability Distribution?

When you focus on a single point prediction, you lose valuable information about the range of possible outcomes. By quantifying uncertainty, you can:
  • Identify best-case and worst-case scenarios
  • Improve risk management and contingency planning
  • Gain confidence in decisions that rely on forecast accuracy

​
What You Will Learn

Quantile Forecasts

Learn how to compute quantile forecasts using TimeGPT.

Prediction Intervals

Discover how to create prediction intervals with TimeGPT.

Forecast Simulation

Generate complete future paths for probability, risk, and scenario analysis.