Forecasting overview

Forecasting is a technique where you analyze historical data in order to make an informed prediction about future trends. For example, you might analyze historical sales data from several store locations in order to predict future sales at those locations. In BigQuery ML, you perform forecasting on time series data.

You can perform forecasting in the following ways:

  • By using the AI.FORECAST function with the built-in TimesFM model. Use this approach when you need to forecast future values for a single variable or multiple variables. This approach doesn't require you to create and manage a model.
  • By using the ML.FORECAST function with the ARIMA_PLUS model. Use this approach when you need to run an ARIMA-based modeling pipeline and decompose the time series into multiple components in order to explain the results. This approach requires you to create and manage a model.
  • By using the ML.FORECAST function with the ARIMA_PLUS_XREG model. Use this approach when you need to forecast future values for multiple variables. This approach requires you to create and manage a model.

In addition to forecasting, you can use the ARIMA_PLUS, ARIMA_PLUS_XREG, and TimesFM models for anomaly detection. For more information, see the following documents:

Compare ARIMA_PLUS models and the TimesFM model

Use the following table to determine whether to use TimesFM, ARIMA_PLUS, or ARIMA_PLUS_XREG model for your use case:

Model type ARIMA_PLUS and ARIMA_PLUS_XREG TimesFM
Model details Statistical model that uses the ARIMA algorithm for the trend component, and a variety of other algorithms for non-trend components. For more information, see Time series modeling pipeline and the publication in the next row. Transformer-based foundation model. For more information, see the publications in the next row.
Publication ARIMA_PLUS: Large-scale, Accurate, Automatic and Interpretable In-Database Time Series Forecasting and Anomaly Detection in BigQuery
Training required Yes, one ARIMA_PLUS or ARIMA_PLUS_XREG model is trained for each time series. No, the TimesFM model is pre-trained.
SQL ease of use High. Requires a CREATE MODEL statement and a function call. Very high. Requires a single function call.
Data history used Uses all time points in the training data, but can be customized to use fewer time points. Can be customized using the context_window parameter of the AI.FORECAST function.
Accuracy Very high. For more information, see publications listed in a previous row. Very high. For more information, see publications listed in a previous row.
Customization High. The CREATE MODEL statement offers arguments that let you tune many model settings, such as the following:
  • Seasonality
  • Holiday effects
  • Step changes
  • Trend
  • Spikes and dips removal
  • Forecasting upper and lower bounds
 Low.
Supports covariates Yes, when using the ARIMA_PLUS_XREG model. Yes, when using the TimesFM 3.0 model (Preview).
Explainability High. You can use the ML.EXPLAIN_FORECAST function to inspect model components. Low.
Model evaluation Use the ML.ARIMA_EVALUATE function. Use the AI.EVALUATE function.
Best use cases
  • You want full control of the model including customization.
  • You need explainability for model output.
  • You want to perform forecasting without creating a model first.
  • You want a high-performance, pre-trained foundational model.

Component analysis

To gain insight into your data and improve forecasting, use time series component analysis to decompose your time series into fundamental underlying patterns. The following functions are available for component analysis and forecasting:

  • ML.SEASONALITY: extracts the seasonality component of a time series, which represents repeating patterns over fixed time periods such as years, weeks, or days.
  • ML.TREND: extracts the trend component of a time series, which represents the directional trajectory of a metric over time and ignores short-term fluctuations or noise.
  • ML.DETECT_CHANGE_POINTS: identifies intervals where the statistical behavior or properties of time series data have shifted. This function provides insight into significant structural breaks, rather than transient or isolated events.

By using the default settings of BigQuery ML's statements and functions, you can create and use a forecasting model even without much ML knowledge. However, having basic knowledge about ML development, and forecasting models in particular, helps you optimize both your data and your model to deliver better results. We recommend using the following resources to develop familiarity with ML techniques and processes: