Proprietary + Confidential

Big Data with

BigQuery

Module 3
Google Cloud Big Data and Machine Learning Fundamentals
 Proprietary + Confidential

Introduction

01
 Proprietary + Confidential

01 Big Data and Machine Learning on Google Cloud

02 Data Engineering for Streaming Data

03 Big Data with BigQuery

04 Machine Learning Options on Google Cloud

05 The Machine Learning Workflow with Vertex AI

In the previous module of this course, you learned how to build a streaming dataflow
using Pub/sub, Dataflow, and Looker vs. Looker Studio. Now let’s switch our focus to
a popular data warehouse product on Google, BigQuery.
 Proprietary + Confidential

Agenda
BigQuery

Storage and analytics

BigQuery demo

BigQuery ML

BigQuery ML

BigQuery ML project phases

BigQuery ML key commands

Hands-on lab

You’ll begin by exploring BigQuery’s two main services, storage and analytics, and
then get a demonstration of BigQuery in use.

After that, you’ll see how BigQuery ML provides a data-to-AI lifecycle all within one
place. You’ll also learn about BigQuery ML project phases, as well as key commands.

Finally, you’ll get hands-on practice using BigQuery ML to build a custom ML model.

Let’s get started!

 Proprietary + Confidential

BigQuery: a fully-managed data warehouse

“Fully managed” means that

BigQuery takes care of the
underlying infrastructure
Terabytes and petabytes
of data gathered from a
wide range of sources

BigQuery is a fully-managed data warehouse.

A data warehouse is a large store, containing terabytes and petabytes of data

gathered from a wide range of sources within an organization, that's used to guide
management decisions.

Being fully managed means that BigQuery takes care of the underlying infrastructure,
so you can focus on using SQL queries to answer business questions–without
worrying about deployment, scalability, and security.

At this point, it’s useful to consider what the main difference is between a data
warehouse and a data lake. A data lake is just a pool of raw, unorganized, and
unclassified data, which has no specified purpose. A data warehouse on the other
hand, contains structured and organized data, which can be used for advanced
querying.
 Proprietary + Confidential

BigQuery features
01
Storage plus analytics
A place to store petabytes of data.
(Petabyte = 11,000 4k movies)

A place to analyze data.

(Machine learning, geospatial analysis, and business intelligence)

Serverless
02 Free from provisioning resources or managing servers but focus on
SQL queries.

03
Flexible pay-as-you-go pricing model
Pay for what your query processes, or use a flat-rate option.

BigQuery
04
Data encryption at rest by default
Encrypt data stored on a disk.

05
Built-in machine learning
Write ML models directly in BigQuery using SQL.

Let’s look at some of the key features of BigQuery.

● BigQuery provides two services in one: storage plus analytics. It’s a place to
store petabytes of data. For reference, 1 petabyte is equivalent to 11,000
movies at 4k quality. BigQuery is also a place to analyze data, with built-in
features like machine learning, geospatial analysis, and business intelligence,
which we’ll explore a bit later on.

● BigQuery is a fully managed serverless solution, meaning that you don’t need
to worry about provisioning any resources or managing servers in the backend
but only focus on using SQL queries to answer your organization's questions in
the frontend. If you’ve never written SQL before, don’t worry. This course
provides resources and labs to help.

● BigQuery has a flexible pay-as-you-go pricing model where you pay for the
number of bytes of data your query processes and for any permanent table
storage. If you prefer to have a fixed bill every month, you can also subscribe
to flat-rate pricing where you have a reserved amount of resources for use.

● Data in BigQuery is encrypted at rest by default without any action required

from a customer. By encryption at rest, we mean encryption used to protect
data that is stored on a disk, including solid-state drives, or backup media.

● BigQuery has built-in machine learning features so you can write ML models
● directly in BigQuery using SQL. Also, if you decide to use other professional
tools—such as Vertex AI from Google Cloud—to train your ML models, you can
export datasets from BigQuery directly into Vertex AI for a seamless
integration across the data-to-AI lifecycle.
 Proprietary + Confidential

Google data warehouse solution architecture

BI tools
● Looker
Real-time data ● Looker Studio
Pub/Sub
(streaming) ● Tableau
● Google Sheets
Dataflow
BigQuery
Cloud
Batch data AI/ML tools
Storage
● AutoML
● Vertex AI
Workbench

So what does a typical data warehouse solution architecture look like?

The input data can be either real-time or batch data. If you think back to the last
module of the course, you’ll recall that there are four challenges of big data in modern
organizations. They are that data can be any format (variety), any size (volume), any
speed (velocity), and possibly inaccurate (veracity).

If it's streaming data, which can be either structured or unstructured, high speed, and
large volume, Pub/Sub is needed to digest the data. If it’s batch data, it can be directly
uploaded to Cloud Storage.

After that, both pipelines lead to Dataflow to process the data. Dataflow is where we
ETL – extract, transform, and load – the data if needed.

BigQuery sits in the middle to link data processes using Dataflow and data access
through analytics, AI, and ML tools.

The job of the analytics engine of BigQuery at the end of a data pipeline is to ingest all
the processed data after ETL, store and analyze it, and possibly output it for further
use such as data visualization and machine learning.

BigQuery outputs usually feed into two buckets: business intelligence tools and AI/ML
tools.
If you’re a business analyst or data analyst, you can connect to visualization tools like
Looker, Looker Studio, Tableau, and other BI tools.

If you prefer to work in spreadsheets, you can query both small or large BigQuery
datasets directly from Google Sheets and even perform common operations like pivot
tables.

Alternatively if you’re a data scientist or machine learning engineer, you can directly
call the data from BigQuery through AutoML or Workbench. These AI/ML tools are
part of Vertex AI, Google's unified ML platform.

BigQuery is like a common staging area for data analytics workloads. When your data
is there, business analysts, BI developers, data scientists, and machine learning
engineers can be granted access to your data for their own insights.
 Proprietary + Confidential

Storage and analytics

02
 Proprietary + Confidential

BigQuery: Storage + analytics

Fully managed storage

facility for datasets

Fast SQL-based
analytical engine

BigQuery provides two services in one. It’s both a fully managed storage facility to
load and store datasets, and also a fast SQL-based analytical engine.

The two services are connected by Google's high-speed internal network. It’s this
super-fast network that allows BigQuery to scale both storage and compute
independently, based on demand.

Let’s look at how BigQuery manages the storage and metadata for datasets.
 Proprietary + Confidential

BigQuery can ingest datasets from various sources

Internal data: Data saved in BigQuery

External data:
○ Google Cloud storage (e.g. Cloud Storage)
○ Google Cloud database (e.g., Spanner or Cloud
SQL)

Multi-cloud data: such as AWS and Azure

Public datasets

Replicated Backed up Autoscaled

BigQuery can ingest datasets from various sources, including:

● internal data, which is data saved directly in BigQuery,

● external data, BigQuery also offers the option to query external data
sources–like data stored in other Google Cloud storage services such as
Cloud Storage, or in other Google Cloud database services, such as Spanner or
Cloud SQL–and bypass BigQuery managed storage. This means a raw CSV file
in Cloud Storage or a Google Sheet can be used to write a query without being
ingested by BigQuery first. One thing to note here: inconsistency might result
from saving and processing data separately. To avoid that risk, consider using
Dataflow to build a streaming data pipeline into BigQuery.
● multi-cloud data, which is data stored in multiple cloud services, such as AWS
or Azure.
● and public datasets. If you don't have data of your own, you can analyze any of
the datasets available in the public dataset marketplace.

After the data is stored in BigQuery, it’s fully managed and is automatically replicated,
backed up, and set up to autoscale.
 Proprietary + Confidential

Patterns to load data into BigQuery

Batch load

Streaming

Generated data
BigQuery

There are three basic patterns to load data into BigQuery.

● The first is a batch load, where source data is loaded into a BigQuery table in a
single batch operation. This can be a one-time operation or it can be automated
to occur on a schedule. A batch load operation can create a new table or append
data into an existing table.
● The second is streaming, where smaller batches of data are streamed
continuously so that the data is available for querying in near-real time.
● And the third is generated data, where SQL statements are used to insert rows
into an existing table or to write the results of a query to a table.
 Proprietary + Confidential

BigQuery can run analytic queries over large datasets

Query terabytes in seconds

Analyze data
Query petabytes in minutes

Of course, the purpose of BigQuery is not to just save data; it’s for analyzing data and
helping to make business decisions.

BigQuery is optimized for running analytic queries over large datasets. It can perform
queries on terabytes of data in seconds and petabytes in minutes. This performance
lets you analyze large datasets efficiently and get insights in near real time.
 Proprietary + Confidential

BigQuery analytics features

Ad hoc analysis

Geospatial analytics

Building machine learning models

Building BI dashboards

Let’s look at the analytics features that are available in BigQuery.

BigQuery supports:

● Ad hoc analysis using Standard SQL, the BigQuery SQL dialect.

● Geospatial analytics using geography data types and Standard SQL geography
functions.
● Building machine learning models using BigQuery ML.
● Building rich, interactive business intelligence dashboards using BigQuery BI
Engine.
 Proprietary + Confidential

BigQuery runs interactive and batch queries

Interactive queries Batch queries

Executed as needed Queries starts when
resources are available

By default, BigQuery runs interactive queries, which means that the queries are
executed as needed.

BigQuery also offers batch queries, where each query is queued on your behalf and
the query starts when idle resources are available, usually within a few minutes.
 Proprietary + Confidential

BigQuery demo

03
 Proprietary + Confidential

Exploring bike share data with BigQuery

Discover the insights of a public dataset using BigQuery.
● Goal: Identify the most popular stations in San Francisco to pick up a shared bike
● Dataset: San Francisco bike share trips
● Schema (partial):

As any data analyst will tell you, exploring a dataset with SQL is often one of the first
steps to uncover hidden insights.

In this section:

● We’ll show you how to use BigQuery to uncover insights from a public dataset.

● The goal is to find the most popular stations across San Francisco to pick up
bikes.

● You can find this public dataset in Bigquery by following these steps:
○ Navigate to the Google Cloud console > BigQuery > Add data > Public
dataset > Search San Francisco bike share > Under
san_francisco_bikeshare, choose bikeshare_trips
○ Click the three dots besides the dataset to start the query, for more
details please refer to the video.

● From the schema, you can find the dataset include information about the:
○ TripID
○ Trip duration
○ Start station, date
○ End station, date
○ Bike number
○ Subscriber information, etc.

● Take a moment to consider this question: How can you find the most popular
station using SQL)?
 Proprietary + Confidential

BigQuery demo

SQL code in BigQuery Query results (partial)

1 SELECT Query complete (1.2 sec elapsed, 81.4 MB processed)

2 start_station_name,
Row start_station_name num_trips
3 COUNT(trip_id) AS num_trips
4 FROM 1 San Francisco Ferry Building 9926
5 `bigquery-public-data.san_francisco_bikeshare.bikeshare_trips`
2 San Francisco Caltrain 9740
6 WHERE start_date > '2017-12-31 00:00:00 UTC'
7 GROUP BY 3 Berry St at 4th St 9041
8 start_station_name
4 Market St at 10th St 8934
9 ORDER BY
10 num_trips DESC
11 LIMIT
12 10

Please find the full demo video here.

The goal of this code is to find the ten most popular stations, based on the number of
trips that start from that station.

● SELECT: The fields to display in the query results

● FROM: Specify the dataset name
● WHERE: The query condition
● GROUP BY: Count the number of trips by start station
● ORDER BY: Order the result
● LIMIT: Specify the number of the query results (top 10 stations in this case)

Partial query results are displayed on the right.

● What insights do you get from these?
● How might you use these insights to improve the business?
● Can you think of any other analysis based on this dataset?
 Proprietary + Confidential

Introduction to

04
BigQuery ML

Although BigQuery started out solely as a data warehouse, over time it has evolved to
provide features that support the data-to-AI lifecycle.
 Proprietary + Confidential

BigQuery ML

Capabilities to build ML models

ML project phases
BigQuery
Key ML commands in SQL

In this section of the course, we’ll explore BigQuery’s capabilities for building machine
learning models and the ML project phases, and walk you through the key ML
commands in SQL.
 Proprietary + Confidential

Building ML models can be time-intensive

Build a model
01 Export data from your datastore into an
integrated development environment (IDE)

02 Transform the data and perform feature

engineering

03 Build the model in TensorFlow and train it

locally or on a virtual machine

To improve model performance, you need to

+ get more data and create new features
Train a model

If you’ve worked with ML models before, you know that building and training them can
be very time-intensive.

● You must first export data from your datastore into an IDE (integrated
development environment) such as Jupyter Notebook or Google Colab and
then transform the data and perform your feature engineering steps before
you can feed it into a training model.
● Then finally, you need to build the model in TensorFlow, or a similar library, and
train it locally on a computer or on a virtual machine.

To improve the model performance, you also need to go back and forth to get more
data and create new features. This process will need to be repeated, but it’s so
time-intensive that you’ll probably stop after a few iterations. Also, I just mentioned
TensorFlow and feature engineering; in the past, if you weren’t familiar with these
technologies, ML was left to the data scientists on your team and was not available to
you.
 Proprietary + Confidential

Step 1: Create a model with a SQL statement

1 CREATE MODEL numbikes.model

2 OPTIONS
3 (model_type='linear_reg', labels=['num_trips']) AS
4 WITH bike_data AS
5 (
6 SELECT
7 COUNT(*) as num_trips,

There are two steps needed to start:

Step 1: Create a model with a SQL statement. Here we can use the bikeshare dataset
as an example.
 Proprietary + Confidential

Step 2: Write a SQL prediction query

and invoke ml.PREDICT

1 SELECT
2 predicted_num_trips,num_trips,trip_data
3 FROM
4 ml.PREDICT(MODEL `numbikes.model`, (WITH bike_data AS
5 (
6 SELECT
7 COUNT(*) as num_trips,

Step 2: Write a SQL prediction query and invoke ml.Predict.

And that’s it! You now have a model and can view the results.

Additional steps might include activities like evaluating the model, but if you know
basic SQL, you can now implement ML; pretty cool!
 Proprietary + Confidential

Hyperparameters
Build a model

Manually control hyperparameters

Automatically tune hyperparameters

Train a model

Hyperparameters

BigQuery ML was designed to be simple, like building a model in two steps. That
simplicity extends to defining the machine learning hyperparameters, which let you
tune the model to achieve the best training result.

Hyperparameters are the settings applied to a model before the training starts, like
the learning rate.

With BigQuery ML, you can either manually control the hyperparameters or hand it to
BigQuery starting with a default hyperparameter setting and then automatic tuning.
 Proprietary + Confidential

ML Models for structured data sets

Supervised models Unsupervised models

Task-driven and identify a goal Data-driven and identify a pattern

Identify patterns
Classify data Predict a number
and clusters
Shoe sales for the next
Is an email spam? Grouping photos
three months

Logistic Linear Cluster

regression regression analysis

When using a structured dataset in BigQuery ML, you need to choose the appropriate
model type. Choosing which type of ML model depends on your business goal and the
datasets.
BigQuery supports supervised models and unsupervised models.

● Supervised models are task-driven and identify a goal.

○ Within a supervised model, if your goal is to classify data, like whether
an email is spam, use logistic regression.
○ If your goal is to predict a number, like shoe sales for the next three
months, use linear regression.
● Alternatively, unsupervised models are data-driven and identify a pattern.
○ Within an unsupervised model, if your goal is to identify patterns or
clusters and then determine the best way to group them, like grouping
random photos of flowers into categories, you should use cluster
analysis.
 Proprietary + Confidential

ML models supported by BigQuery ML

Logistic regression k-means clustering

Classification Other models
DNN classifier (TensorFlow) Time series forecasting (ARIMA+)

XGBoost Recommendation: Matrix factorization

AutoML Tables
Anomaly Detection
Wide and Deep NNs

Linear regression
Regression ML Ops Importing TensorFlow models for batch
DNN regressor (TensorFlow) prediction

XGBoost Exporting models from BigQuery

ML for online prediction
AutoML Tables
Hyperparameter tuning using
Wide and Deep NNs Vertex AI Vizier

Once you have your problem outlined, it’s time to decide on the best model.
Categories include classification and regression models. There are also other model
options to choose from, along with ML ops.

Logistic regression is an example of a classification model, and linear regression is an

example of a regression model. We recommend that you start with these options,
and use the results to benchmark to compare against more complex models such as
DNN (deep neural networks), which may take more time and computing resources to
train and deploy.
 Proprietary + Confidential

Machine learning operations

Importing TensorFlow models

Machine
learning Operations
development
Exporting models from BigQuery ML

Hyperparameter tuning
ML Ops

In addition to providing different types of machine learning models, BigQuery ML

supports features to deploy, monitor, and manage the ML production, called ML Ops,
which is short for machine learning operations.

Options include:
● Importing TensorFlow models for batch prediction
● Exporting models from BigQuery ML for online prediction
● And hyperparameter tuning using Vertex AI Vizier

We’ll explore ML Ops in more detail later in this course.

 Proprietary + Confidential

05 Using BigQuery ML to
predict customer
lifetime value
 Proprietary + Confidential

Predict customer lifetime value with a model

Estimate how much revenue or

profit you can expect from a
customer given their history and
customers with similar patterns

Lifetime value (LTV)

Now that you’re familiar with the types of ML models available to choose from,
high-quality data must be used to teach the models what they need to learn. The best
way to learn the key concepts of machine learning on structured datasets is through
an example.

In this scenario, we’ll predict customer lifetime value with a model.

Lifetime value, or LTV, is a common metric in marketing used to estimate how much
revenue or profit you can expect from a customer given their history and customers
with similar patterns.
 Proprietary + Confidential

Google Analytics ecommerce data set

hits_product_V2ProductName

177 categories

Google Men’s Vintage Badge Tee Black

Google Women’s 1/4 Zip Performance Pullover Two-Tone Blue

Google Men’s Lightweight Microfleece Jacket Black

Google Women’s Quilted Insulated Vest Black

Google Women’s Lightweight Microfleece Jacket

Identify high-value customers
Women’s Weatherblock Shell Jacket Black

Men’s Weatherblock Shell Jacket Black

We’ll use a Google Analytics ecommerce dataset from Google’s own merchandise
store that sells branded items like t-shirts and jackets.

The goal is to identify high-value customers and bring them to our store with special
promotions and incentives.
 Proprietary + Confidential

Fields
Customer Total Average time Total Ecommerce
lifetime visits spent on revenue transactions
pageviews the site

Row fullVisitorID distinct_days_visited ltv_pageviews ltv_visits ltv_avg_time_on_site_s ltv_revenue ltv_transactions label

1 7813149964104484438 79 7395 138 479.63 624572 67

2 7713012430069756739 2 514 6 1954.33 18194 35 High value customer

3 6760732402251466726 30 868 41 723.55 481282 34

4 552667592603848032 1 466 1 7013.0 8796 25 High value customer

5 1957458976293878100 148 4303 284 796.46 7711343 22

6 4983264713004875783 2 366 4 3807.5 7485 21 High value customer

7 2402527199731150932 28 559 31 906.61 327010 19

Having explored the available fields, you may find some useful in determining whether
a customer is high value based on their behavior on our website.

These fields include:

● customer lifetime pageviews,
● total visits,
● average time spent on the site,
● total revenue brought in,
● and ecommerce transactions on the site.

Remember that in machine learning, you feed in columns of data and let the model
figure out the relationship to best predict the label. It may even turn out that some of
the columns weren’t useful at all to the model in predicting the outcome -- you’ll see
later how to determine this.
 Proprietary + Confidential

Labels

Row fullVisitorID distinct_days_visited ltv_pageviews ltv_visits ltv_avg_time_on_site_s ltv_revenue ltv_transactions label

1 7813149964104484438 79 7395 138 479.63 624572 67

2 7713012430069756739 2 514 6 1954.33 18194 35 High value customer

3 6760732402251466726 30 868 41 723.55 481282 34

4 552667592603848032 1 466 1 7013.0 8796 25 High value customer

5 1957458976293878100 148 4303 284 796.46 7711343 22

6 4983264713004875783 2 366 4 3807.5 7485 21 High value customer

7 2402527199731150932 28 559 31 906.61 327010 19

Example / Observation / Instance Label Label

Linear regression Logistic regression

Now that we have some data, we can prepare to feed it into the model. Incidentally, to
keep this example simple, we’re only using seven records, but we’d need tens of
thousands of records to train a model effectively.

Before we feed the data into the model, we first need to define our data and columns
in the language that data scientists and other ML professionals use.

Using the Google Merchandise Store example, a record or row in the dataset is called
an example, an observation, or an instance.

A label is a correct answer, and you know it’s correct because it comes from historical
data. This is what you need to train the model on in order to predict future data.
Depending on what you want to predict, a label can be either a numeric variable, which
requires a linear regression model, or a categorical variable, which requires a logistic
regression model.

For example, if we know that a customer who has made transactions in the past and
spends a lot of time on our website often turns out to have high lifetime revenue, we
could use revenue as the label and predict the same for newer customers with that
same spending trajectory.

This means forecasting a number, so we can use a linear regression as a starting

point to model.
Labels could also be categorical variables like binary values, such as High Value
Customer or not. To predict a categorical variable, if you recall from the previous
section, you need to use a logistic regression model.

Knowing what you’re trying to predict, such as a class or a number, will greatly
influence the type of model you’ll use.
 Proprietary + Confidential

Features

Row fullVisitorID distinct_days_visited ltv_pageviews ltv_visits ltv_avg_time_on_site_s ltv_revenue ltv_transactions label

1 7813149964104484438 79 7395 138 479.63 624572 67

2 7713012430069756739 2 514 6 1954.33 18194 35 High value customer

3 6760732402251466726 30 868 41 723.55 481282 34

4 552667592603848032 1 466 1 7013.0 8796 25 High value customer

5 1957458976293878100 148 4303 284 796.46 7711343 22

6 4983264713004875783 2 366 4 3807.5 7485 21 High value customer

7 2402527199731150932 28 559 31 906.61 327010 19

Features

Sifting through data can be time consuming

But what do we call all the other data columns in the data table?

Those columns are called features, or potential features. Each column of data is like a
cooking ingredient you can use from the kitchen pantry. Too many ingredients,
however, can ruin a dish!

The process of sifting through data can be time consuming. Understanding the quality
of the data in each column and working with teams to get more features or more
history is often the hardest part of any ML project.
 Proprietary + Confidential

Feature engineering

Row fullVisitorID distinct_days_visited ltv_pageviews ltv_visits ltv_avg_time_on_site_s ltv_revenue ltv_transactions label

1 7813149964104484438 79 7395 138 479.63 624572 67

2 7713012430069756739 2 514 6 1954.33 18194 35 High value customer

3 6760732402251466726 30 868 41 723.55 481282 34

4 552667592603848032 1 466 1 7013.0 8796 25 High value customer

5 1957458976293878100 148 4303 284 796.46 7711343 22

6 4983264713004875783 2 366 4 3807.5 7485 21 High value customer

7 2402527199731150932 28 559 31 906.61 327010 19

Feature
engineering

Calculate fields

You can even combine or transform feature columns in a process called feature
engineering.

If you’ve ever created calculated fields in SQL, you’ve already executed the basics of
feature engineering.
 Proprietary + Confidential

BigQuery ML does much of the hard work

01 Automatically one-hot encoding categorical values

02 Automatically splits the dataset into training data

and evaluation data

BigQuery ML

BigQuery ML does much of the hard work for you, like automatically one-hot encoding
categorical values. One-hot encoding is a method of converting categorical data to
numeric data to prepare it for model training. From there, BigQuery ML automatically
splits the dataset into training data and evaluation data.
 Proprietary + Confidential

Predicting on future data

Historical Training Data (Known LTV)
Row fullVisitorID distinct_days_visited ltv_pageviews ltv_visits ltv_avg_time_on_site_s ltv_revenue ltv_transactions label

1 7813149964104484438 79 7395 138 479.63 624572 67

2 7713012430069756739 2 514 6 1954.33 18194 35 High value customer

3 6760732402251466726 30 868 41 723.55 481282 34

4 552667592603848032 1 466 1 7013.0 8796 25 High value customer

5 1957458976293878100 148 4303 284 796.46 7711343 22

6 4983264713004875783 2 366 4 3807.5 7485 21 High value customer

7 2402527199731150932 28 559 31 906.61 327010 19

Future Data (Unknown LTV)

8 7904807859681747547 3 42 3 1162.0 null null ??????????????????

9 4405445121320750966 51 358 62 517.36 null null ??????????????????

10 1419607020881916790 5 22 5 711.0 null null ??????????????????

And finally, there is predicting on future data.

Let’s say new data comes in that you don’t have a label for, so you don’t know whether
it is for a high-value customer. You do, however, have a rich history of labeled
examples for you to train a model on.
 Proprietary + Confidential

Predicting results!
Historical Training Data (Known LTV)
Row fullVisitorID distinct_days_visited ltv_pageviews ltv_visits ltv_avg_time_on_site_s ltv_revenue ltv_transactions label

1 7813149964104484438 79 7395 138 479.63 624572 67

2 7713012430069756739 2 514 6 1954.33 18194 35 High value customer

3 6760732402251466726 30 868 41 723.55 481282 34

4 552667592603848032 1 466 1 7013.0 8796 25 High value customer

5 1957458976293878100 148 4303 284 796.46 7711343 22

6 4983264713004875783 2 366 4 3807.5 7485 21 High value customer

7 2402527199731150932 28 559 31 906.61 327010 19

Future Data (Unknown LTV)

8 7904807859681747547 3 42 3 1162.0 null null High value customer

9 4405445121320750966 51 358 62 517.36 null null

10 1419607020881916790 5 22 5 711.0 null null High value customer

So if we train a model on the known historical data, and are happy with the
performance, then we can use it to predict on future datasets!
 Proprietary + Confidential

BigQuery ML

06
project phases
 Proprietary + Confidential

Phase 1 and 2: Prepare data

01
● Easy connectors to Google products.
Extract, transform, and load data into BigQuery
● Use SQL joins.

● Use SQL to create the training dataset.

02 Select and preprocess features
● BigQuery ML does preprocessing for you.

03 Create the model inside BigQuery

04 Evaluate the performance of the trained model

05 Use the model to make predictions

Let’s explore the key phases of a machine learning project.

In phase 1, you extract, transform, and load data into BigQuery, if it isn’t there
already. If you’re already using other Google products, like YouTube for example, look
out for easy connectors to get that data into BigQuery before you build your own
pipeline. You can enrich your existing data warehouse with other data sources by
using SQL joins.

In phase 2, you select and preprocess features. You can use SQL to create the
training dataset for the model to learn from. You’ll recall that BigQuery ML does some
of the preprocessing for you, like one-hot encoding of your categorical variables.
One-hot encoding converts your categorical data into numeric data that is required by
a training model.
 Proprietary + Confidential

Phase 3: Create ML model

01 Extract, transform, and load data into BigQuery Use the “CREATE MODEL” command.

#standardSQL
02 Select and preprocess features CREATE MODEL
ecommerce.classification

OPTIONS
03 Create the model inside BigQuery
(
model_type='logistic_reg',
input_label_cols =
04 Evaluate the performance of the trained model 'will_buy_later'
) AS

# SQL query with training data

05 Use the model to make predictions

In phase 3, you create the model inside BigQuery. This is done by using the “CREATE
MODEL” command. Give it a name, specify the model type, and pass in a SQL query
with your training dataset. From there, you can run the query.
 Proprietary + Confidential

Phase 4: Evaluate ML model

01 Extract, transform, and load data into BigQuery

Execute an ML.EVALUATE query.

02 Select and preprocess features

#standardSQL
SELECT
roc_auc,
03 Create the model inside BigQuery
accuracy,
precision,
recall

04 Evaluate the performance of the trained model FROM

ML.EVALUATE(MODEL
`ecommerce.classification`)

05 Use the model to make predictions # SQL query with evaluation data

In phase 4, after your model is trained, you can execute an ML.EVALUATE query to
evaluate the performance of the trained model on your evaluation dataset. It’s here
that you can analyze loss metrics like a Root Mean Squared Error for forecasting
models and area-under-the-curve, accuracy, precision, and recall, for classification
models. We’ll explore these metrics later in the course.
 Proprietary + Confidential

Phase 5: Make prediction

01 Extract, transform, and load data into BigQuery

02 Select and preprocess features

Invoke the ml.PREDICT command.

03 Create the model inside BigQuery
#standardSQL
SELECT * FROM
04 Evaluate the performance of the trained model ML.PREDICT
(MODEL
`ecommerce.classification`)

05 Use the model to make predictions # SQL query with test data

In phase 5, the final phase, when you’re happy with your model performance, you can
then use it to make predictions. To do so, invoke the ml.PREDICT command on your
newly trained model to return with predictions and the model’s confidence in those
predictions. With the results, your label field will have “predicted” added to the field
name. This is your model’s prediction for that label.
 Proprietary + Confidential

BigQuery ML
key commands

07
Now that you’re familiar with the key phases of an ML project, let’s explore some of
the key commands of BigQuery ML.
 Proprietary + Confidential

Create a model with CREATE MODEL

CREATE OR REPLACE MODEL

`mydataset.mymodel`
OPTIONS
( model_type='linear_reg',
input_label_cols= 'sales'
ls_init_learn_rate=.15,
l1_reg=1,
max_iterations=5 ) AS

You’ll remember from an earlier that you can create a model with just the CREATE
MODEL command.

If you want to overwrite an existing model, use the CREATE OR REPLACE MODEL
command.

Models have OPTIONS, which you can specify. The most important, and the only one
required, is the model type.
 Proprietary + Confidential

Inspect what a model learned with ML.WEIGHTS

SELECT
category, Output
weight
FROM Each feature has a
weight from -1 to 1.
UNNEST((
SELECT
Closer to 0 =
category_weights the less important the
feature is to the
FROM prediction.
ML.WEIGHTS(MODEL `bracketology.ncaa_model`)
WHERE Closer to -1 or 1 =
the more important
processed_input = 'seed')) # try other features the feature is to the
like 'school_ncaa' prediction.

ORDER BY weight DESC

You can inspect what a model learned with the ML.WEIGHTS command and filtering
on an input column.

The output of ML.WEIGHTS is a numerical value, and each feature has a weight from
-1 to 1. That value indicates how important the feature is for predicting the result, or
label. If the number is closer to 0, the feature isn't important for the prediction.
However, if the number is closer to -1 or 1, then the feature is more important for
predicting the result.
 Proprietary + Confidential

Evaluate the model's performance with ML.EVALUATE

SELECT
*
FROM
ML.EVALUATE(MODEL `bracketology.ncaa_model`)

To evaluate the model's performance, you can run an ML.EVALUATE command

against a trained model. You get different performance metrics depending on the
model type you chose.
 Proprietary + Confidential

Make batch predictions with ML.PREDICT

CREATE OR REPLACE TABLE `bracketology.predictions` AS (

SELECT * FROM ML.PREDICT(MODEL `bracketology.ncaa_model`,

# predicting for 2018 tournament games (2017 season)

(SELECT * FROM
`data-to-insights.ncaa.2018_tournament_results`)
)
)

And if you want to make batch predictions, you can use the ML.PREDICT command on
a trained model, and pass through the dataset you want to make the prediction on.
 Proprietary + Confidential

BigQuery ML commands for supervised models

Labels Identify column as ‘label’ or specify column in OPTIONS using input_label_cols.

Data columns that are part of your SELECT statement, after your CREATE MODEL statement.
Features
SELECT * FROM ML.FEATURE_INFO(MODEL `mydataset.mymodel`)

Model object An object created in BigQuery that resides in your BigQuery dataset.

Linear Regression, logistic Regression, etc.

Model types CREATE OR REPLACE MODEL <dataset>.<name>

OPTIONS(model_type='<type>') AS
<training dataset>

Training progress SELECT * FROM ML.TRAINING_INFO(MODEL `mydataset.mymodel`)

Inspect weights SELECT * FROM ML.WEIGHTS(MODEL `mydataset.mymodel`, (<query>))

Evaluation SELECT * FROM ML.EVALUATE(MODEL `mydataset.mymodel`)

Prediction SELECT * FROM ML.PREDICT(MODEL `mydataset.mymodel`, (<query>))

Now let’s explore a consolidated list of BigQuery ML commands for supervised

models.
● First in BigQuery ML, you need a field in your training dataset titled LABEL, or
you need to specify which field, or fields, your labels are using as the input
label columns in your model OPTIONS.
● Second, your model features are the data columns that are part of your
SELECT statement after your CREATE MODEL statement. After a model is
trained, you can use the ML.FEATURE_INFO command to get statistics and
metrics about that column for additional analysis.
● Next is the model object itself. This is an object created in BigQuery that
resides in your BigQuery dataset. You train many different models, which will
all be objects stored under your BigQuery dataset, much like your tables and
views. Model objects can display information for when it was last updated or
how many training runs it completed.
● Creating a new model is as easy as writing CREATE MODEL, choosing a type,
and passing in a training dataset. Again, if you’re predicting on a numeric field,
such as next year's sales, consider linear regression for forecasting. If it's a
discrete class like high, medium, low, or spam/not-spam, consider using
logistic regression for classification.
● While the model is running, and even after it’s complete, you can view training
progress with ML.TRAINING_INFO.
● As mentioned earlier, you can inspect weights to see what the model learned
about the importance of each feature as it relates to the label you’re
predicting. The importance is indicated by the weight of each feature.
● You can see how well the model performed against its evaluation dataset by
using ML.EVALUATE.
● And lastly, getting predictions is as simple as writing ML.PREDICT and
referencing your model name and prediction dataset.
 Proprietary + Confidential

Lab:

08
Predicting purchases
with BigQuery ML
 Proprietary + Confidential

Hands-on lab

Order data

Visitor data

BigQuery ML

shop.googlemerchandisestore.com

Now it’s time to get some hands-on practice building a machine learning model in
BigQuery.

In the next lab, you’ll use ecommerce data from the Google Merchandise Store
website: https://shop.googlemerchandisestore.com/

The site’s visitor and order data have been loaded into BigQuery, and you’ll build a
machine learning model to predict whether a visitor will return for more purchases
later.
 Proprietary + Confidential

You’ll practice

1 Loading data using BigQuery

2 Querying and exploring the dataset

3 Creating a training and evaluation dataset

4 Creating a classification model

5 Evaluating model performance

6 Predicting and ranking purchase probability

You’ll get practice:

● Loading data into BigQuery from a public dataset.

● Querying and exploring the ecommerce dataset.
● Creating a training and evaluation dataset to be used for batch prediction.
● Creating a classification (logistic regression) model in BigQueryML.
● Evaluating the performance of your machine learning model.
● And predicting and ranking the probability that a visitor will make a purchase
 Proprietary + Confidential

09 Summary

Well done on completing another lab! Hopefully you now feel more comfortable
building custom machine learning models with BigQuery ML!

Let’s review what we explored in this module of the course.

 Proprietary + Confidential

Storage and analytics

Fully-managed storage
facility for datasets

Fast SQL-based
BigQuery analytical engine

Our focus was on BigQuery, the data warehouse that provides two services in one. It’s
a fully-managed storage facility for datasets, and a fast SQL-based analytical engine.
 Proprietary + Confidential

BigQuery sits between data processes and data uses

BI tools
● Looker
Real-time data ● Looker Studio
Pub/Sub
(streaming) ● Tableau
● Google Sheets
Dataflow
BigQuery
Cloud
Batch data AI/ML tools
Storage
● AutoML
● Vertex AI
Workbench

BigQuery sits between data processes and data uses, like a common staging area. It
gets data from ingestion and processing and outputs data to BI tools such as Looker
and Looker Studio and ML tools such as Vertex AI.

After the data is in BigQuery, business analysts, BI developers, data scientists, and
machine learning engineers can be granted access to the data for their own insights.
 Proprietary + Confidential

Key phases of a BigQuery ML project

01 Extract, transform, and load data into BigQuery

02 Select and preprocess features

03 Create the model inside BigQuery

BigQuery ML
04 Evaluate the performance of the trained model

05 Use the model to make predictions

In addition to traditional data warehouses, BigQuery offers machine learning features.

This means you can use BigQuery to directly build ML models in five key phases.

In phase 1, you extract, transform, and load data into BigQuery, if it isn’t there
already.

In phase 2, you select and preprocess features. You can use SQL to create the
training dataset for the model to learn from.

In phase 3, you create the ML model inside BigQuery.

In phase 4, after your model is trained, you can execute an ML.EVALUATE query to
evaluate the performance of the trained model on your evaluation dataset.

And in phase 5, the final phase, when you’re happy with your model performance, you
can then use it to make predictions.
 Proprietary + Confidential

Quiz
 Proprietary + Confidential

Question #1
Question

BigQuery is a fully managed data warehouse. What does “fully managed” refer to?
A. BigQuery manages the data quality for you.
B. BigQuery manages the underlying structure for you.
C. BigQuery manages the cost for you.
D. BigQuery manages the data source for you.
 Proprietary + Confidential

Question #1
Answer

BigQuery is a fully managed data warehouse. What does “fully managed” refer to?
A. BigQuery manages the data quality for you.
B. BigQuery manages the underlying structure for you.
C. BigQuery manages the cost for you.
D. BigQuery manages the data source for you.
 Proprietary + Confidential

Question #2
Question

Which two services does BigQuery provide?

A. Storage and compute
B. Application services and analytics
C. Application services and storage
D. Storage and analytics
 Proprietary + Confidential

Question #2
Answer

Which two services does BigQuery provide?

A. Storage and compute
B. Application services and analytics
C. Application services and storage
D. Storage and analytics
 Proprietary + Confidential

Question #3
Question

Which pattern describes source data that is moved into a BigQuery table in a single
operation?
A. Batch load
B. Streaming
C. Spot load
D. Generated data
 Proprietary + Confidential

Question #3
Answer

Which pattern describes source data that is moved into a BigQuery table in a single
operation?
A. Batch load
B. Streaming
C. Spot load
D. Generated data
 Proprietary + Confidential

Question #4
Question

Which BigQuery feature leverages geography data types and standard SQL geography
functions to analyze a data set?
A. Ad hoc analysis
B. Geospatial analysis
C. Building machine learning models
D. Building business intelligence dashboards
 Proprietary + Confidential

Question #4
Answer

Which BigQuery feature leverages geography data types and standard SQL geography
functions to analyze a data set?
A. Ad hoc analysis
B. Geospatial analysis
C. Building machine learning models
D. Building business intelligence dashboards
 Proprietary + Confidential

Question #5
Question

Data is loaded into BigQuery, and the features have been selected and preprocessed.
What’s next when using BigQuery ML?
A. Use the ML model to make predictions.
B. Create the ML model inside BigQuery.
C. Evaluate the performance of the trained ML model.
D. Classify labels to train on historical data.
 Proprietary + Confidential

Question #5
Answer

Data is loaded into BigQuery, and the features have been selected and preprocessed.
What’s next when using BigQuery ML?
A. Use the ML model to make predictions.
B. Create the ML model inside BigQuery.
C. Evaluate the performance of the trained ML model.
D. Classify labels to train on historical data.
 Proprietary + Confidential

Question #6
Question

In a supervised machine learning model, what provides historical data that can be used
to predict future data?
A. Labels
B. Examples
C. Data points
D. Features
 Proprietary + Confidential

Question #6
Answer

In a supervised machine learning model, what provides historical data that can be used
to predict future data?
A. Labels
B. Examples
C. Data points
D. Features
 Proprietary + Confidential

Question #7
Question

You want to use machine learning to group random photos into similar groups. Which
should you use?
A. Unsupervised learning, cluster analysis
B. Supervised learning, linear regression
C. Supervised learning, logistic regression
D. Unsupervised learning, dimensionality reduction
 Proprietary + Confidential

Question #7
Answer

You want to use machine learning to group random photos into similar groups. Which
should you use?
A. Unsupervised learning, cluster analysis
B. Supervised learning, linear regression
C. Supervised learning, logistic regression
D. Unsupervised learning, dimensionality reduction
 Proprietary + Confidential

Question #8
Question

You want to use machine learning to identify whether an email is spam. Which should
you use?
A. Unsupervised learning, cluster analysis
B. Supervised learning, linear regression
C. Supervised learning, logistic regression
D. Unsupervised learning, dimensionality reduction
 Proprietary + Confidential

Question #8
Answer

You want to use machine learning to identify whether an email is spam. Which should
you use?
A. Unsupervised learning, cluster analysis
B. Supervised learning, linear regression
C. Supervised learning, logistic regression
D. Unsupervised learning, dimensionality reduction