Google Updates BigQuery with Multi-Time Series Forecasting and Enhanced Security Features

Google has introduced updates to BigQuery, enhancing both its machine learning capabilities and security features.

New Multi-Time Series Forecasting in BigQuery ML

Google has launched a preview of multi-time series forecasting in BigQuery ML through the new TIME_SERIES_ID_COL option. This feature is now available in ARIMA_PLUS_XREG multivariate time series models.

Key Technical Specifications

The TIME_SERIES_ID_COL option supports multiple data types including:

• STRING
• INT64
• ARRAY<STRING>
• ARRAY<INT64>

The forecasting horizon can be configured using the HORIZON parameter, which:

• Defaults to 1,000 time points
• Has a maximum value of 10,000 time points
• Requires an INT64 value input

Security Enhancement with VPC Allowlists

In a separate update, Google has made BigQuery Omni Virtual Private Cloud (VPC) allowlists generally available. This security feature enables organizations to:

• Restrict access to AWS S3 buckets
• Control access to Azure Blob Storage
• Manage permissions from specific BigQuery Omni VPCs