Supported input feature types

BigQuery ML supports different input feature types for different model types. Supported input feature types are listed in the following table:

Model Category	Model Types	Numeric types (INT64, NUMERIC, BIGNUMERIC, FLOAT64)	Categorical types (BOOL, STRING, BYTES, DATE, DATETIME)	TIMESTAMP	STRUCT	GEOGRAPHY	ARRAY<Numeric types>	ARRAY<Categorical types>	ARRAY<STRUCT<INT64, Numeric types>>
Supervised Learning	Linear & Logistic Regression	✔	✔	✔	✔		✔	✔	✔
	Deep Neural Networks	✔	✔		✔		✔	✔	✔
	Wide-and-Deep	✔	✔		✔		✔	✔	✔
	Boosted trees	✔	✔		✔		✔	✔	✔
	AutoML Tables	✔	✔	✔	✔		✔	✔
Unsupervised Learning	K-means	✔	✔	✔	✔	✔	✔	✔
	PCA	✔	✔	✔	✔		✔	✔
	Autoencoder	✔	✔	✔	✔		✔	✔	✔
Time Series Models	ARIMA_PLUS_XREG	✔	✔	✔	✔			✔	✔

Note: Matrix Factorization and ARIMA_PLUS models have special input feature types. The input types listed for ARIMA_PLUS_XREG are only for external regressors.

Dense vector input

BigQuery ML supports ARRAY<numerical> as dense vector input during model training. The embedding feature is a special type of dense vector. see the ML.GENERATE_EMBEDDING function for more information.

Sparse input

BigQuery ML supports ARRAY<STRUCT> as sparse input during model training. Each struct contains an INT64 value that represents its zero-based index, and a numeric type that represents the corresponding value.

Below is an example of a sparse tensor input for the integer array [0,1,0,0,0,0,1]:

ARRAY<STRUCT<k INT64, v INT64>>[(1, 1), (6, 1)] AS f1