End-to-end user journey for each model

BigQuery ML supports a variety of machine learning models and a complete machine learning flow for each model, such as feature preprocessing, model creation, hyperparameter tuning, inference, evaluation, and model export. The machine learning flow for the models are split into the following two tables:

Model creation phase
Model use phase

Model creation phase

Model category	Model types	Model creation	Feature preprocessing	Hyperparameter tuning	Model weights	Feature & training info	Tutorials
Supervised learning	Linear & logistic regression	create model	Automatic preprocessing, Manual preprocessing¹	HP tuning² ml.trial_info	ml.weights	ml.feature_info ml.training_info	regression classification
	Deep neural networks (DNN)	create model			N/A⁵		N/A
	Wide & Deep networks	create model			N/A⁵		N/A
	Boosted trees	create model			N/A⁵		N/A
	Random forest	create model			N/A⁵		N/A
	AutoML classification & regression	create model	N/A³	N/A³	N/A⁵		N/A
Unsupervised learning	K-means	create model	Automatic preprocessing, Manual preprocessing¹	HP tuning² ml.trial_info	ml.centroids	ml.feature_info ml.training_info	cluster bike stations
	Matrix factorization	create model	N/A	HP tuning² ml.trial_info	ml.weights		explicit feedback implicit feedback
	Principal component analysis (PCA)	create model	Automatic preprocessing, Manual preprocessing¹	N/A	ml.principal_ components, ml.principal_ component_info		N/A
	Autoencoder	create model	Automatic preprocessing, Manual preprocessing¹	HP tuning² ml.trial_info	N/A⁵		N/A
Time series models	ARIMA_PLUS	create model	Automatic preprocessing	auto.ARIMA⁴	ml.arima_ coefficients	ml.feature_info ml.training_info	single series multi-series millions of series custom holidays limit forecasted values hierarchical time series forecasting
Time series models	ARIMA_PLUS_XREG	create model	Automatic preprocessing	auto.ARIMA⁴	ml.arima_ coefficients	ml.feature_info ml.training_info	multivariate forecasting
Generative AI remote models	Remote model over a Vertex AI text generation model⁶	create model	N/A	N/A	N/A	N/A	generate text with remote model use tuning and evaluation to improve LLM performance
Generative AI remote models	Remote model over a Vertex AI embedding generation model⁶	create model	N/A	N/A	N/A	N/A	generate text embeddings with a remote model generate image embeddings with a remote model generate and search multimodal embeddings
AI remote models	Remote model over the Cloud Vision API	create model	N/A	N/A	N/A	N/A	N/A
	Remote model over the Cloud Translation API	create model	N/A	N/A	N/A	N/A	N/A
	Remote model over the Cloud Natural Language API	create model	N/A	N/A	N/A	N/A	N/A
	Remote model over the Document AI API (Preview)	create model	N/A	N/A	N/A	N/A	N/A
	Remote model over the Speech-to-Text API (Preview)	create model	N/A	N/A	N/A	N/A	N/A
Remote models	Remote model with a Vertex AI endpoint	create model	N/A	N/A	N/A	N/A	predict with remote model
Imported models	TensorFlow	create model	N/A	N/A	N/A	N/A	predict with imported TensorFlow model
	TensorFlow Lite	create model	N/A	N/A	N/A	N/A	N/A
	Open Neural Network Exchange (ONNX)	create model	N/A	N/A	N/A	N/A	Predict with scikit-learn models in ONNX format Predict with PyTorch models in ONNX format
	XGBoost	create model	N/A	N/A	N/A	N/A	N/A
Transform-only models⁷	Transform-only (Preview)	create model	Manual preprocessing¹	N/A	N/A	ml.feature_info	N/A

¹See TRANSFORM clause for the feature engineering tutorial. For more information about the preprocessing functions, see the BQML - Feature Engineering Functions tutorial.

²See use hyperparameter tuning to improve model performance tutorial.

³Automatic feature engineering and hyperparameter tuning are embedded in the AutoML model training by default.

⁴The auto.ARIMA algorithm performs hyperparameter tuning for the trend module. Hyperparameter tuning is not supported for the entire modeling pipeline. See the modeling pipeline for more details.

⁵BigQuery ML doesn't support functions that retrieve the weights for boosted trees, random forest, DNNs, Wide-and-deep, Autoencoder, or AutoML models. To see the weights of those models, you can export an existing model from BigQuery ML to Cloud Storage and then use the XGBoost library or the TensorFlow library to visualize the tree structure for the tree models or the graph structure for the neural networks. For more information, see the EXPORT MODEL documentation and the EXPORT MODEL tutorial.

⁶Uses a Vertex AI foundation model or customizes it by using supervised tuning.

⁷This is not a typical ML model but rather an artifact that transforms raw data into features.

Model use phase

Model category	Model types	Evaluation	Inference	AI explanation	Model monitoring	Model export	Tutorials
Supervised learning	Linear & logistic regression	ml.evaluate ml.confusion_matrix¹ ml.roc_curve²	ml.predict ml.transform	ml.explain_predict³ ml.global_explain ml.advanced_weights⁸	ml.describe_data ml.validate_data_drift ml.validate_data_skew ml.tfdv_describe ml.tfdv_validate	export model⁵	regression classification
	Deep neural networks (DNN)						N/A
	Wide & Deep networks						N/A
	Boosted trees			ml.explain_predict³ ml.global_explain ml.feature_importance⁴			N/A
	Random forest						N/A
	AutoML classification & regression		ml.predict	ml.global_explain	ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate		N/A
Unsupervised learning	K-means	ml.evaluate	ml.predict ml.detect_anomalies ml.transform	N/A	ml.describe_data ml.validate_data_drift ml.validate_data_skew ml.tfdv_describe ml.tfdv_validate	export model⁵	cluster bike stations
	Matrix factorization		ml.recommend ml.generate_embedding (Preview)		ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate		explicit feedback implicit feedback
	Principal component analysis (PCA)		ml.predict ml.generate_embedding (Preview) ml.detect_anomalies ml.transform		ml.describe_data ml.validate_data_drift ml.validate_data_skew ml.tfdv_describe ml.tfdv_validate		N/A
	Autoencoder		ml.predict ml.generate_embedding (Preview) ml.detect_anomalies ml.reconstruction_loss ml.transform				N/A
Time series models	ARIMA_PLUS	ml.evaluate ml.arima_evaluate⁶ ml.holiday_info	ml.forecast ml.detect_anomalies	ml.explain_forecast⁷	ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate	N/A	single series multi-series millions of series custom holidays limit forecasted values hierarchical time series forecasting
Time series models	ARIMA_PLUS_XREG	ml.evaluate ml.arima_evaluate⁶ ml.holiday_info	ml.forecast ml.detect_anomalies (Preview)	ml.explain_forecast⁷	ml.describe_data ml.validate_data_drift ml.validate_data_skew ml.tfdv_describe ml.tfdv_validate	N/A	multivariate forecasting
Generative AI remote models	Remote model over a Vertex AI text generation model⁹	ml.evaluate¹¹ (Preview)	ml.generate_text	N/A	ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate	N/A	generate text with remote model use tuning and evaluation to improve LLM performance
Generative AI remote models	Remote model over a Vertex AI embedding generation model⁹	N/A	ml.generate_embedding	N/A		N/A	generate text embeddings with a remote model generate image embeddings with a remote model generate and search multimodal embeddings
AI remote models	Remote model over the Cloud Vision API	N/A	ml.annotate_image	N/A	ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate	N/A	N/A
	Remote model over the Cloud Translation API	N/A	ml.translate	N/A		N/A	N/A
	Remote model over the Cloud Natural Language API	N/A	ml.understand_text	N/A		N/A	N/A
	Remote model over the Document AI API (Preview)	N/A	ml.process_document	N/A		N/A	N/A
	Remote model over the Speech-to-Text API (Preview)	N/A	ml.transcribe	N/A		N/A	N/A
Remote models	Remote model with a Vertex AI endpoint	N/A	ml.predict	N/A	ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate	N/A	predict with remote model
Imported models	TensorFlow	N/A	ml.predict	N/A	ml.describe_data ml.validate_data_drift ml.tfdv_describe ml.tfdv_validate	export model⁵	predict with imported TensorFlow model
	TensorFlow Lite	N/A	ml.predict	N/A		N/A	N/A
	Open Neural Network Exchange (ONNX)	N/A	ml.predict	N/A		N/A	Predict with scikit-learn models in ONNX format Predict with PyTorch models in ONNX format
	XGBoost	N/A	ml.predict	N/A		N/A	N/A
Transform-only models¹⁰ (Preview)	Transform-only (Preview)	N/A	ml.transform	N/A	N/A	export model⁵	N/A

¹ml.confusion_matrix is only applicable to classification models.

²ml.roc_curve is only applicable to binary classification models.

³ml.explain_predict is an extended version of ml.predict. For more information, see Explainable AI overview. To learn how ml.explain_predict is used, see regression tutorial and classification tutorial.

⁴For the difference between ml.global_explain and ml.feature_importance, see Explainable AI overview.

⁵See the Export a BigQuery ML model for online prediction tutorial. For more information about online serving, see the BQML - Create Model with Inline Transpose tutorial.

⁶For ARIMA_PLUS or ARIMA_PLUS_XREG models, ml.evaluate can take new data as input to compute forecasting metrics such as mean absolute percentage error (MAPE). In the absence of new data, ml.evaluate has an extended version ml.arima_evaluate which outputs different evaluation information.

⁷ml.explain_forecast is an extended version of ml.forecast. For more information, see Explainable AI overview. To learn how ml.explain_forecast is used, see the visualize results steps of the single time series forecasting and multiple time series forecasting tutorials.

⁸ml.advanced_weights is an extended version of ml.weights, see ml.advanced_weights for more details.

⁹Uses a Vertex AI foundation model or customizes it by using supervised tuning.

¹⁰This is not a typical ML model but rather an artifact that transforms raw data into features.

¹¹Not supported for all Vertex AI LLMs. For more information, see ml.evaluate.