Use the Google Cloud console to create a tabular dataset and train a classification model.
Create a tabular dataset
In the Google Cloud console, in the Vertex AI section, go to the Datasets page.
Click Create in the button bar to create a new dataset.
Structured_AutoML_Tutorialfor the dataset name and select the Tabular tab.
Select the Regression/Classification objective.
Leave the Region set to us-central1.
Click Create to create the dataset.
For this tutorial, you'll use a publicly available bank dataset hosted on Cloud Storage.
For Select a data source, click Select CSV files from Cloud Storage
In Import file path, enter
Analyze the dataset
The analyze section lets you view more information about the dataset, like missing or NULL values.
Because our dataset is formatted correctly for this tutorial, you don't need to do anything on this page and can skip this section.
Optional. Click Generate statistics to view the number of missing or NULL values in the dataset. This can take 10 minutes or longer.
Optional. Click on one of the feature columns to learn more about the data values.
Train an AutoML classification model
Click Train new model.
In the Training method pane, confirm that the dataset you created previously is selected for the Dataset field.
For the Objective field, select Classification.
Confirm that the AutoML training method is selected.
In the Model details pane, select Deposit for the target column and click Continue.
The target column is what we're training the model to predict. For the
Depositcolumn indicates whether the client purchased a term deposit (2 = yes, 1 = no).
The Training options pane gives you an opportunity to add features and transform column data. If no columns are selected, then by default all non- target columns will be used as features for training. This dataset is ready to use, so there's no need to apply any transformations.
In the Compute and pricing pane, enter
1for the training budget.
The training budget determines actual training time, but the time to complete training includes other activities, so the entire process can take longer than one hour. When the model finishes training, it is displayed in the model tab as a live link, with a green checkmark status icon.
Click Start training.
Your model is now being trained, which can take an hour or more to complete. You'll receive an email when training is complete. When your model has finished training, follow the next page of this tutorial to deploy your model and request a prediction.
This tutorial uses a dataset that's been cleaned and formatted for AutoML training, but most data will require some work before it's ready to be used. The quality of your training data impacts the effectiveness of the models you create. Learn more about preparing data.
Sourcing and preparing your data is a critical to ensuring an accurate machine learning model. Learn more about best practices.
Learn more about creating a tabular dataset.
Vertex AI offers two model training methods, AutoML and custom training. AutoML lets you train with minimal effort and machine learning experience, while custom training gives you complete control over training functionality. Learn more about training methods.
Vertex AI examines the source data type and feature values and infers how it will use that feature in model training. It's recommended that you review each column's data type to verify that it's been interpreted correctly. If needed, you can specify a different supported transformation for any feature. Learn more about transformations.
Learn more about training an AutoML for classification or regression.