In this tutorial, you use a linear regression model in BigQuery ML to predict the weight of a penguin based on the penguin's demographic information. A linear regression is a type of regression model that generates a continuous value from a linear combination of input features.
This tutorial uses the
bigquery-public-data.ml_datasets.penguins
dataset.
Objectives
In this tutorial, you will perform the following tasks:
- Create a linear regression model.
- Evaluate the model.
- Make predictions by using the model.
Costs
This tutorial uses billable components of Google Cloud, including the following:
- BigQuery
- BigQuery ML
For more information on BigQuery costs, see the BigQuery pricing page.
For more information on BigQuery ML costs, see BigQuery ML pricing.
Before you begin
-
In the Google Cloud console, on the project selector page, select or create a Google Cloud project.
-
Make sure that billing is enabled for your Google Cloud project.
-
Enable the BigQuery API.
Required permissions
To create the model using BigQuery ML, you need the following IAM permissions:
bigquery.jobs.create
bigquery.models.create
bigquery.models.getData
bigquery.models.updateData
bigquery.models.updateMetadata
To run inference, you need the following permissions:
bigquery.models.getData
on the modelbigquery.jobs.create
Create a dataset
Create a BigQuery dataset to store your ML model:
In the Google Cloud console, go to the BigQuery page.
In the Explorer pane, click your project name.
Click
View actions > Create dataset.On the Create dataset page, do the following:
For Dataset ID, enter
bqml_tutorial
.For Location type, select Multi-region, and then select US (multiple regions in United States).
The public datasets are stored in the
US
multi-region. For simplicity, store your dataset in the same location.Leave the remaining default settings as they are, and click Create dataset.
Create the model
Create a linear regression model using the Analytics sample dataset for BigQuery.
SQL
You can create a linear regression model by using the
CREATE MODEL
statement
and specifying LINEAR_REG
for the model type. Creating the model includes
training the model.
The following are useful things to know about the CREATE MODEL
statement:
- The
input_label_cols
option specifies which column in theSELECT
statement to use as the label column. Here, the label column isbody_mass_g
. For linear regression models, the label column must be real-valued, that is, the column values must be real numbers. This query's
SELECT
statement uses the following columns in thebigquery-public-data.ml_datasets.penguins
table to predict a penguin's weight:species
: the species of penguin.island
: the island that the penguin resides on.culmen_length_mm
: the length of the penguin's culmen in millimeters.culmen_depth_mm
: the depth of the penguin's culmen in millimeters.flipper_length_mm
: the length of the penguin's flippers in millimeters.sex
: the sex of the penguin.
The
WHERE
clause in this query'sSELECT
statement,WHERE body_mass_g IS NOT NULL
, excludes rows where thebody_mass_g
column isNULL
.
Run the query that creates your linear regression model:
In the Google Cloud console, go to the BigQuery page.
In the query editor, run the following query:
CREATE OR REPLACE MODEL `bqml_tutorial.penguins_model` OPTIONS (model_type='linear_reg', input_label_cols=['body_mass_g']) AS SELECT * FROM `bigquery-public-data.ml_datasets.penguins` WHERE body_mass_g IS NOT NULL;
It takes about 30 seconds to create the
penguins_model
model. To see the model, go to the Explorer pane, expand thebqml_tutorial
dataset, and then expand the Models folder.
BigQuery DataFrames
Before trying this sample, follow the BigQuery DataFrames setup instructions in the BigQuery quickstart using BigQuery DataFrames. For more information, see the BigQuery DataFrames reference documentation.
To authenticate to BigQuery, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.
It takes about 30 seconds to create the model. To see the model, go to the Explorer
pane, expand the bqml_tutorial
dataset, and then expand the Models folder.
Get training statistics
To see the results of the model training, you can use the
ML.TRAINING_INFO
function,
or you can view the statistics in the Google Cloud console. In this
tutorial, you use the Google Cloud console.
A machine learning algorithm builds a model by examining many examples and attempting to find a model that minimizes loss. This process is called empirical risk minimization.
Loss is the penalty for a bad prediction. It is a number indicating how bad the model's prediction was on a single example. If the model's prediction is perfect, the loss is zero; otherwise, the loss is greater. The goal of training a model is to find a set of weights and biases that have low loss, on average, across all examples.
See the model training statistics that were generated when you ran the
CREATE MODEL
query:
In the Explorer pane, expand the
bqml_tutorial
dataset and then the Models folder. Click penguins_model to open the model information pane.Click the Training tab, and then click Table. The results should look similar to the following:
The Training Data Loss column represents the loss metric calculated after the model is trained on the training dataset. Since you performed a linear regression, this column shows the mean squared error value. A normal_equation optimization strategy is automatically used for this training, so only one iteration is required to converge to the final model. For more information on setting the model optimization strategy, see
optimize_strategy
.
Evaluate the model
After creating the model, evaluate the model's performance by using the
ML.EVALUATE
function or the score
BigQuery DataFrames function to evaluate the predicted values generated by the model against the actual data.
SQL
For input, the ML.EVALUATE
function takes the trained model and a dataset
that matches the schema of the data that you used to train the model. In
a production environment, you should
evaluate the model on different data than the data you used to train the model.
If you run ML.EVALUATE
without providing input data, the function retrieves
the evaluation metrics calculated during training. These metrics are calculated
by using the automatically reserved evaluation dataset:
SELECT
*
FROM
ML.EVALUATE(MODEL bqml_tutorial.penguins_model
);
Run the ML.EVALUATE
query:
In the Google Cloud console, go to the BigQuery page.
In the query editor, run the following query:
SELECT * FROM ML.EVALUATE(MODEL `bqml_tutorial.penguins_model`, ( SELECT * FROM `bigquery-public-data.ml_datasets.penguins` WHERE body_mass_g IS NOT NULL));
BigQuery DataFrames
Before trying this sample, follow the BigQuery DataFrames setup instructions in the BigQuery quickstart using BigQuery DataFrames. For more information, see the BigQuery DataFrames reference documentation.
To authenticate to BigQuery, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.
The results should look similar to the following:
Because you performed a linear regression, the results include the following columns:
mean_absolute_error
mean_squared_error
mean_squared_log_error
median_absolute_error
r2_score
explained_variance
An important metric in the evaluation results is the
R2 score.
The R2 score is a statistical measure that determines if the linear
regression predictions approximate the actual data. A value of 0
indicates
that the model explains none of the variability of the response data around the
mean. A value of 1
indicates that the model explains all the variability of
the response data around the mean.
You can also look at the model's information pane in the Google Cloud console to view the evaluation metrics:
Use the model to predict outcomes
Now that you have evaluated your model, the next step is to use it to predict
an outcome. You can run the
ML.PREDICT
function
on the model to predict the body mass in grams of all penguins that reside on
the Biscoe Islands.
For input, the ML.PREDICT
function takes the trained model and a dataset that
matches the schema of the data that you used to train the model, excluding the
label column.
Run the ML.PREDICT
query:
In the Google Cloud console, go to the BigQuery page.
In the query editor, run the following query:
SELECT * FROM ML.PREDICT(MODEL `bqml_tutorial.penguins_model`, ( SELECT * FROM `bigquery-public-data.ml_datasets.penguins` WHERE island = 'Biscoe'));
The results should look similar to the following:
Explain the prediction results
To understand why the model is generating these prediction results, you can use
the
ML.EXPLAIN_PREDICT
function.
ML.EXPLAIN_PREDICT
is an extended version of the ML.PREDICT
function.
ML.EXPLAIN_PREDICT
not only outputs prediction results, but also outputs
additional columns to explain the prediction results. In practice, you can run
ML.EXPLAIN_PREDICT
instead of ML.PREDICT
. For more information, see
BigQuery ML explainable AI overview.
Run the ML.EXPLAIN_PREDICT
query:
In the Google Cloud console, go to the BigQuery page.
In the query editor, run the following query:
SELECT * FROM ML.EXPLAIN_PREDICT(MODEL `bqml_tutorial.penguins_model`, ( SELECT * FROM `bigquery-public-data.ml_datasets.penguins` WHERE island = 'Biscoe'), STRUCT(3 as top_k_features));
The results should look similar to the following:
For linear regression models, Shapley values are used to generate feature
attribution values for each feature in the model. ML.EXPLAIN_PREDICT
outputs
the top three feature attributions per row of the penguins
table because
top_k_features
was set to 3
in the query. These attributions are sorted by
the absolute value of the attribution in descending order. In all examples, the
feature sex
contributed the most to the overall prediction.
Globally explain the model
To know which features are generally the most important to determine penguin
weight, you can use the
ML.GLOBAL_EXPLAIN
function.
In order to use ML.GLOBAL_EXPLAIN
, you must retrain the model with the
ENABLE_GLOBAL_EXPLAIN
option set to TRUE
.
Retrain and get global explanations for the model:
In the Google Cloud console, go to the BigQuery page.
In the query editor, run the following query to retrain the model:
#standardSQL CREATE OR REPLACE MODEL `bqml_tutorial.penguins_model` OPTIONS ( model_type = 'linear_reg', input_label_cols = ['body_mass_g'], enable_global_explain = TRUE) AS SELECT * FROM `bigquery-public-data.ml_datasets.penguins` WHERE body_mass_g IS NOT NULL;
In the query editor, run the following query to get global explanations:
SELECT * FROM ML.GLOBAL_EXPLAIN(MODEL `bqml_tutorial.penguins_model`)
The results should look similar to the following:
Clean up
To avoid incurring charges to your Google Cloud account for the resources used in this tutorial, either delete the project that contains the resources, or keep the project and delete the individual resources.
- You can delete the project you created.
- Or you can keep the project and delete the dataset.
Delete your dataset
Deleting your project removes all datasets and all tables in the project. If you prefer to reuse the project, you can delete the dataset you created in this tutorial:
If necessary, open the BigQuery page in the Google Cloud console.
In the navigation, click the bqml_tutorial dataset you created.
Click Delete dataset on the right side of the window. This action deletes the dataset, the table, and all the data.
In the Delete dataset dialog box, confirm the delete command by typing the name of your dataset (
bqml_tutorial
) and then click Delete.
Delete your project
To delete the project:
- In the Google Cloud console, go to the Manage resources page.
- In the project list, select the project that you want to delete, and then click Delete.
- In the dialog, type the project ID, and then click Shut down to delete the project.
What's next
- For an overview of BigQuery ML, see Introduction to BigQuery ML.
- For information on creating models, see the
CREATE MODEL
syntax page.