内置 TimesFM 单变量模型是 Google Research 的开源 TimesFM 模型的实现。Google Research TimesFM 模型是一种时序预测基础模型,已通过许多真实世界数据集中的数十亿个时间点进行预训练,因此您可以将其应用于许多领域的新预测数据集。所有 BigQuery 支持的区域都提供 TimesFM 模型。
[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-08-17。"],[],[],null,["# The TimesFM model\n=================\n\n|\n| **Preview**\n|\n|\n| This product or feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA products and features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n| **Note:** To give feedback or request support for this feature, contact [bqml-feedback@google.com](mailto:bqml-feedback@google.com).\n\nThis document describes BigQuery ML's built-in\nTimesFM time series forecasting model.\n\nThe built-in TimesFM univariate model is an implementation of Google Research's\nopen source\n[TimesFM model](https://github.com/google-research/timesfm). The Google Research\nTimesFM model is a foundation model for time-series forecasting that has been\npre-trained on billions of time-points from many real-world datasets, so you\ncan apply it to new forecasting datasets across many domains.\nThe TimesFM model is available in all BigQuery supported regions.\n\nUsing BigQuery ML's built-in TimesFM model with the\n[`AI.FORECAST` function](/bigquery/docs/reference/standard-sql/bigqueryml-syntax-ai-forecast)\nlets you perform\nforecasting without having to create and train your own model, so you can\navoid the need for model management.\nThe forecast results from the TimesFM model are comparable to\nconventional statistical methods such as ARIMA. If you want more\nmodel tuning options than the TimesFM model offers, you can create an\n[`ARIMA_PLUS`](/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-time-series)\nor\n[`ARIMA_PLUS_XREG`](/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-multivariate-time-series)\nmodel and use it with the\n[`ML.FORECAST` function](/bigquery/docs/reference/standard-sql/bigqueryml-syntax-forecast)\ninstead.\n\nTo try using a TimesFM model with the `AI.FORECAST` function, see\n[Forecast multiple time series with a TimesFM univariate model](/bigquery/docs/timesfm-time-series-forecasting-tutorial).\n\nTo learn more about the Google Research TimesFM model, use the following\nresources:\n\n- [Google Research blog](https://research.google/blog/a-decoder-only-foundation-model-for-time-series-forecasting/)\n- [GitHub repository](https://github.com/google-research/timesfm)\n- [Hugging Face page](https://huggingface.co/collections/google/timesfm-release-66e4be5fdb56e960c1e482a6)"]]