The TimesFM model
This document describes BigQuery ML's built-in TimesFM time series forecasting model.
The built-in TimesFM model is an implementation of Google Research's open source TimesFM model. The Google Research TimesFM model is a foundation model for time-series forecasting that has been pre-trained on billions of time-points from many real-world datasets, so you can apply it to new forecasting datasets across many domains. The TimesFM model is available in all BigQuery supported regions.
Using BigQuery ML's built-in TimesFM model with the
AI.FORECAST function
lets you perform
forecasting without having to create and train your own model, so you can
avoid the need for model management.
The forecast results from the TimesFM model are comparable to
conventional statistical methods such as ARIMA. If you want more
model tuning options than the TimesFM model offers, you can create an
ARIMA_PLUS
or
ARIMA_PLUS_XREG
model and use it with the
ML.FORECAST function
instead.
To learn how to use a TimesFM model with the AI.FORECAST function, see the
following tutorials:
- Forecast single or multiple time series with a TimesFM univariate model
- Forecast a single time series with a TimesFM multivariate model (Preview)
- Forecast multiple time series with a TimesFM multivariate model (Preview)
To use the TimesFM model to detect anomalies in time series data, use the
AI.DETECT_ANOMALIES function.
To evaluate forecasted values from the TimesFM model against the actual values,
use the
AI.EVALUATE function.
To learn more about the Google Research TimesFM model, use the following resources:
When you use TimesFM through BigQuery, your usage is governed by
the Google Cloud Terms of Service and allows
for commercial uses. The non-commercial license associated with the publicly
downloadable TimesFM 3.0 weights on GitHub and Hugging Face applies only to
self-hosted downloads and does not restrict usage within
BigQuery.