Agent Platform lets you perform machine learning with tabular data using simple processes and interfaces. You can create the following model types for your tabular data problems:
- Binary classificationmodels predict a binary outcome (one of two classes). Use this model type for yes or no questions. For example, you might want to build a binary classification model to predict whether a customer would buy a subscription. Generally, a binary classification problem requires less data than other model types.
- Multi-class classificationmodels predict one class from three or more discrete classes. Use this model type for categorization. For example, as a retailer, you might want to build a multi-class classification model to segment customers into different personas.
- Regressionmodels predict a continuous value. For example, as a retailer, you might want to build a regression model to predict how much a customer will spend next month.
- Forecastingmodels predict a sequence of values. For example, as a retailer, you might want to forecast daily demand of your products for the next 3 months so that you can appropriately stock product inventories in advance.
For an introduction to machine learning with tabular data, see Introduction to Tabular Data . For further information about Agent Platform solutions, see Agent Platform solutions for classification and regression and Agent Platform solutions for forecasting .
A note about fairness
Google commits to making progress in following responsible AI practices . To this end, our ML products, including AutoML, are designed around core principles such as fairness and human-centered machine learning . For more information about best practices for mitigating bias when building your own ML system, see Inclusive ML guide - AutoML .
Agent Platform solutions for classification and regression
Agent Platform offers the following solutions for classification and regression:
Tabular Workflow for End-to-End AutoML
Tabular Workflow for End-to-End AutoML is a complete AutoML pipeline for classification and regression tasks. It is similar to the AutoML API , but allows you to choose what to control and what to automate. Instead of having controls for the whole pipeline, you have controls for every step in the pipeline. These pipeline controls include:
- Data splitting
- Feature engineering
- Architecture search
- Model training
- Model ensembling
- Model distillation
Benefits
- Supports large datasetsthat are multiple TB in size and have up to 1000 columns.
- Allows you to improve stability and lower training timeby limiting the search space of architecture types or skipping architecture search.
- Allows you to improve training speedby manually selecting the hardware used for training and architecture search.
- Allows you to reduce model size and improve latencywith distillation or by changing the ensemble size.
- Each AutoML component can be inspected in a powerful pipelines graph interface that lets you see the transformed data tables, evaluated model architectures, and many more details.
- Each AutoML component gets extended flexibility and transparency, such as being able to customize parameters, hardware, view process status, logs, and more.
To learn more about Tabular Workflows, see Tabular Workflows on Agent Platform . To learn more about Tabular Workflow for End-to-End AutoML, see Tabular Workflow for End-to-End AutoML .
Classification and regression with AutoML
Agent Platform offers integrated, fully managed pipelines for end-to-end classification or regression tasks. Agent Platform searches for the optimal set of hyperparameters, trains multiple models with multiple sets of hyperparameters, and then creates a single, final model from an ensemble of the top models. Agent Platform considers neural networks and boosted trees for the model types.
Benefits
- Easy to use: Agent Platform chooses the model type, model parameters, and hardware for you.
For further information, see Classification and Regression Overview .
Agent Platform solutions for forecasting
Agent Platform offers the following solutions for forecasting:
- Tabular Workflow for Forecasting
- Forecasting with AutoML
- Forecasting with BigQuery ML ARIMA_PLUS
- Forecasting with Prophet
Tabular Workflow for Forecasting
Tabular Workflow for Forecasting is the complete pipeline for forecasting tasks. It is similar to the AutoML API , but lets you to choose what to control and what to automate. Instead of having controls for the whole pipeline, you have controls for every step in the pipeline. These pipeline controls include:
- Data splitting
- Feature engineering
- Architecture search
- Model training
- Model ensembling
Benefits
- Supports large datasetsthat are up to 1TB in size and have up to 200 columns.
- Lets you improve stability and lower training timeby limiting the search space of architecture types or skipping architecture search.
- Lets you improve training speedby manually selecting the hardware used for training and architecture search.
- Lets you reduce model size and improve latencyby changing the ensemble size.
- Each component can be inspected in a powerful pipelines graph interface that lets you see the transformed data tables, evaluated model architectures and many more details.
- Each component gets extended flexibility and transparency, such as being able to customize parameters, hardware, view process status, logs and more.
To learn more about Tabular Workflows, see Tabular Workflows on Agent Platform . To learn more about Tabular Workflow for Forecasting, see Tabular Workflow for Forecasting .
Forecasting with AutoML
Agent Platform offers an integrated, fully managed pipeline for end-to-end forecasting tasks. Agent Platform searches for the optimal set of hyperparameters, trains multiple models with multiple sets of hyperparameters, and then creates a single, final model from an ensemble of the top models. You can choose between Time series Dense Encoder (TiDE) , Temporal Fusion Transformer (TFT) , AutoML (L2L) , and Seq2Seq+ for your model training method. Agent Platform considers only neural networks for the model type.
Benefits
- Easy to use: Agent Platform chooses model parameters and hardware for you.
For further information, see Forecasting Overview .
Forecasting with BigQuery ML ARIMA_PLUS
BigQuery ML ARIMA_PLUS is a univariate forecasting model. As a statistical model, it is faster to train than a model based on neural networks . We recommend training a BigQuery ML ARIMA_PLUS model if you need to perform many quick iterations of model training or if you need an inexpensive baseline to measure other models against.
Like Prophet , BigQuery ML ARIMA_PLUS attempts to decompose each time series into trends, seasons, and holidays, producing a forecast using the aggregation of these models' inferences. One of the many differences, however, is that BQML ARIMA+ uses ARIMA to model the trend component, while Prophet attempts to fit a curve using a piecewise logistic or linear model.
Google Cloud offers a pipeline for training a BigQuery ML ARIMA_PLUS model and a pipeline for getting batch inferences from a BigQuery ML ARIMA_PLUS model. Both pipelines are instances of Vertex AI Pipelines from Google Cloud Pipeline Components (GCPC).
Benefits
- Easy to use: BigQuery chooses model parameters and hardware for you.
- Fast: model training provides a low-cost baseline to compare other models against.
For further information, see Forecasting with ARIMA+ .
Forecasting with Prophet
Prophet is a forecasting model maintained by Meta. See the Prophet paper for algorithm details and the documentation for more information about the library.
Like BigQuery ML ARIMA_PLUS , Prophet attempts to decompose each time series into trends, seasons, and holidays, producing a forecast using the aggregation of these models' inferences. An important difference, however, is that BQML ARIMA+ uses ARIMA to model the trend component, while Prophet attempts to fit a curve using a piecewise logistic or linear model.
Google Cloud offers a pipeline for training a Prophet model and a pipeline for getting batch inferences from a Prophet model. Both pipelines are instances of Vertex AI Pipelines from Google Cloud Pipeline Components (GCPC).
Integration of Prophet with Agent Platform means that you can do the following:
- Use Agent Platform data splitting and windowing strategies .
- Read data from either BigQuery tables or CSVs stored in Cloud Storage. Agent Platform expects each row to have the same format as Agent Platform Forecasting .
Although Prophet is a multivariate model, Agent Platform supports only a univariate version of it.
Benefits
- Flexible: you can improve training speed by selecting the hardware used for training
For further information, see Forecasting with Prophet .
What's next
- Learn about machine learning with tabular data .
- Learn about classification and regression with AutoML .
- Learn about forecasting with AutoML .
- Learn about forecasting with Prophet .
- Learn about forecasting with BigQuery ML ARIMA_PLUS .
- Learn about Tabular Workflows .
