3.9 Forecasting

3.9.1 Forecasting Model Types

3.9.2 Training a Forecasting Model

3.9.3 Using a Forecasting Model

3.9.4 Prediction Intervals

To generate a forecast, run ML_TRAIN and specify a forecasting task as a JSON object literal. ML_PREDICT_TABLE can then predict the values for the selected column. Use ML_SCORE to score the model quality.

Create a timeseries forecast for a single column with a numeric data type. In forecasting terms, this is a univariate endogenous variable.

HeatWave AutoML supports multivariate endogenous forecasting models, and exogenous forecasting models.