What Is Azure Machine Learning Studio? | Microsoft Docs Manual Azura

User Manual:

Open the PDF directly: View PDF PDF.
Page Count: 421 [warning: Documents this large are best viewed by clicking the View PDF Link!]

Table of ContentsTable of Contents
Machine Learning Studio Documentation
Overview
Machine Learning Studio
ML Studio capabilities
ML Studio basics (infographic)
Frequently asked questions
What's new?
Get Started
Create your first experiment
Example walkthrough
Create a predictive solution
1 - Create a workspace
2 - Upload data
3 - Create experiment
4 - Train and evaluate
5 - Deploy web service
6 - Access web service
Data Science for Beginners
1 - Five questions
2 - Is your data ready?
3 - Ask the right question
4 - Predict an answer
5 - Copy other people's work
R quick start
How To
Set up tools and utilities
Manage a workspace
Acquire and understand data
Import training data
Develop models
Create and train models
Operationalize models
Overview
Deploy models
Manage web services
Retrain models
Consume models
Examples
Sample experiments
Sample datasets
Customer churn example
Reference
Azure PowerShell module (New)
Azure PowerShell module (Classic)
Algorithm & Module reference
REST management APIs
Web service error codes
Related
Azure AI Gallery
Overview
Industries
Solutions
Experiments
Jupyter Notebooks
Competitions
Competitions FAQ
Tutorials
Collections
Custom Modules
Resources
Azure Roadmap
Net# Neural Networks Language
Pricing
Service updates
Blog
MSDN forum
Stack Overflow
Videos
What is Azure Machine Learning Studio?
4/9/2018 • 9 min to read • Edit Online
NOTENOTE
The Machine Learning Studio interactive workspace
TIPTIP
Microsoft Azure Machine Learning Studio is a collaborative, drag-and-drop tool you can use to build, test, and
deploy predictive analytics solutions on your data. Machine Learning Studio publishes models as web services that
can easily be consumed by custom apps or BI tools such as Excel.
Machine Learning Studio is where data science, predictive analytics, cloud resources, and your data meet.
You can try Azure Machine Learning for free. No credit card or Azure subscription is required. Get started now.
To develop a predictive analysis model, you typically use data from one or more sources, transform and analyze
that data through various data manipulation and statistical functions, and generate a set of results. Developing a
model like this is an iterative process. As you modify the various functions and their parameters, your results
converge until you are satisfied that you have a trained, effective model.
Azure Machine Learning Studio gives you an interactive, visual workspace to easily build, test, and iterate on a
predictive analysis model. You drag-and-drop datasets and analysis modules onto an interactive canvas,
connecting them together to form an experiment, which you run in Machine Learning Studio. To iterate on your
model design, you edit the experiment, save a copy if desired, and run it again. When you're ready, you can convert
your training experiment to a predictive experiment, and then publish it as a web service so that your model
can be accessed by others.
There is no programming required, just visually connecting datasets and modules to construct your predictive
analysis model.
To download and print a diagram that gives an overview of the capabilities of Machine Learning Studio, see Overview
diagram of Azure Machine Learning Studio capabilities.
Get started with Machine Learning Studio
Cortana IntelligenceCortana Intelligence
Azure Machine Learning StudioAzure Machine Learning Studio
GalleryGallery
Components of an experiment
When you first enter Machine Learning Studio you see the Home page. From here you can view documentation,
videos, webinars, and find other valuable resources.
Click the upper-left menu and you'll see several options.
Click Cortana Intelligence and you'll be taken to the home page of the Cortana Intelligence Suite. The Cortana
Intelligence Suite is a fully managed big data and advanced analytics suite to transform your data into intelligent
action. See the Suite home page for full documentation, including customer stories.
There are two options here, Home, the page where you started, and Studio.
Click Studio and you'll be taken to the Azure Machine Learning Studio. First you'll be asked to sign in using
your Microsoft account, or your work or school account. Once signed in, you'll see the following tabs on the left:
PROJECTS - Collections of experiments, datasets, notebooks, and other resources representing a single project
EXPERIMENTS - Experiments that you have created and run or saved as drafts
WEB SERVICES - Web services that you have deployed from your experiments
NOTEBOOKS - Jupyter notebooks that you have created
DATASETS - Datasets that you have uploaded into Studio
TRAINED MODELS - Models that you have trained in experiments and saved in Studio
SETTINGS - A collection of settings that you can use to configure your account and resources.
Click Gallery and you'll be taken to the Azure AI Gallery. The Gallery is a place where a community of data
scientists and developers share solutions created using components of the Cortana Intelligence Suite.
For more information about the Gallery, see Share and discover solutions in the Azure AI Gallery.
DatasetsDatasets
ModulesModules
Deploying a predictive analytics web service
An experiment consists of datasets that provide data to analytical modules, which you connect together to
construct a predictive analysis model. Specifically, a valid experiment has these characteristics:
The experiment has at least one dataset and one module
Datasets may be connected only to modules
Modules may be connected to either datasets or other modules
All input ports for modules must have some connection to the data flow
All required parameters for each module must be set
You can create an experiment from scratch, or you can use an existing sample experiment as a template. For more
information, see Copy example experiments to create new machine learning experiments.
For an example of creating a simple experiment, see Create a simple experiment in Azure Machine Learning
Studio.
For a more complete walkthrough of creating a predictive analytics solution, see Develop a predictive solution with
Azure Machine Learning.
A dataset is data that has been uploaded to Machine Learning Studio so that it can be used in the modeling
process. A number of sample datasets are included with Machine Learning Studio for you to experiment with, and
you can upload more datasets as you need them. Here are some examples of included datasets:
MPG data for various automobiles - Miles per gallon (MPG) values for automobiles identified by number of
cylinders, horsepower, etc.
Breast cancer data - Breast cancer diagnosis data.
Forest fires data - Forest fire sizes in northeast Portugal.
As you build an experiment you can choose from the list of datasets available to the left of the canvas.
For a list of sample datasets included in Machine Learning Studio, see Use the sample data sets in Azure Machine
Learning Studio.
A module is an algorithm that you can perform on your data. Machine Learning Studio has a number of modules
ranging from data ingress functions to training, scoring, and validation processes. Here are some examples of
included modules:
Convert to ARFF - Converts a .NET serialized dataset to Attribute-Relation File Format (ARFF).
Compute Elementary Statistics - Calculates elementary statistics such as mean, standard deviation, etc.
Linear Regression - Creates an online gradient descent-based linear regression model.
Score Model - Scores a trained classification or regression model.
As you build an experiment you can choose from the list of modules available to the left of the canvas.
A module may have a set of parameters that you can use to configure the module's internal algorithms. When you
select a module on the canvas, the module's parameters are displayed in the Properties pane to the right of the
canvas. You can modify the parameters in that pane to tune your model.
For some help navigating through the large library of machine learning algorithms available, see How to choose
algorithms for Microsoft Azure Machine Learning.
Once your predictive analytics model is ready, you can deploy it as a web service right from Machine Learning
Studio. For more details on this process, see Deploy an Azure Machine Learning web service.
Key machine learning terms and concepts
Data exploration, descriptive analytics, and predictive analyticsData exploration, descriptive analytics, and predictive analytics
Supervised and unsupervised learningSupervised and unsupervised learning
Model training and evaluationModel training and evaluation
Training dataTraining data
Evaluation dataEvaluation data
Other common machine learning terms
Machine learning terms can be confusing. Here are definitions of key terms to help you. Use comments following
to tell us about any other term you'd like defined.
Data exploration is the process of gathering information about a large and often unstructured data set in order
to find characteristics for focused analysis.
Data mining refers to automated data exploration.
Descriptive analytics is the process of analyzing a data set in order to summarize what happened. The vast
majority of business analytics - such as sales reports, web metrics, and social networks analysis - are descriptive.
Predictive analytics is the process of building models from historical or current data in order to forecast future
outcomes.
Supervised learning algorithms are trained with labeled data - in other words, data comprised of examples of the
answers wanted. For instance, a model that identifies fraudulent credit card use would be trained from a data set
with labeled data points of known fraudulent and valid charges. Most machine learning is supervised.
Unsupervised learning is used on data with no labels, and the goal is to find relationships in the data. For
instance, you might want to find groupings of customer demographics with similar buying habits.
A machine learning model is an abstraction of the question you are trying to answer or the outcome you want to
predict. Models are trained and evaluated from existing data.
When you train a model from data, you use a known data set and make adjustments to the model based on the
data characteristics to get the most accurate answer. In Azure Machine Learning, a model is built from an
algorithm module that processes training data and functional modules, such as a scoring module.
In supervised learning, if you're training a fraud detection model, you use a set of transactions that are labeled as
either fraudulent or valid. You split your data set randomly, and use part to train the model and part to test or
evaluate the model.
Once you have a trained model, evaluate the model using the remaining test data. You use data you already know
the outcomes for, so that you can tell whether your model predicts accurately.
algorithm: A self-contained set of rules used to solve problems through data processing, math, or automated
reasoning.
anomaly detection: A model that flags unusual events or values and helps you discover problems. For
example, credit card fraud detection looks for unusual purchases.
categorical data: Data that is organized by categories and that can be divided into groups. For example a
categorical data set for autos could specify year, make, model, and price.
classification: A model for organizing data points into categories based on a data set for which category
groupings are already known.
feature engineering: The process of extracting or selecting features related to a data set in order to enhance
the data set and improve outcomes. For instance, airfare data could be enhanced by days of the week and
holidays. See Feature selection and engineering in Azure Machine Learning.
module: A functional part in a Machine Learning Studio model, such as the Enter Data module that enables
Next steps
entering and editing small data sets. An algorithm is also a type of module in Machine Learning Studio.
model: A supervised learning model is the product of a machine learning experiment comprised of training
data, an algorithm module, and functional modules, such as a Score Model module.
numerical data: Data that has meaning as measurements (continuous data) or counts (discrete data). Also
referred to as quantitative data.
partition: The method by which you divide data into samples. See Partition and Sample for more information.
prediction: A prediction is a forecast of a value or values from a machine learning model. You might also see
the term "predicted score." However, predicted scores are not the final output of a model. An evaluation of the
model follows the score.
regression: A model for predicting a value based on independent variables, such as predicting the price of a car
based on its year and make.
score: A predicted value generated from a trained classification or regression model, using the Score Model
module in Machine Learning Studio. Classification models also return a score for the probability of the
predicted value. Once you've generated scores from a model, you can evaluate the model's accuracy using the
Evaluate Model module.
sample: A part of a data set intended to be representative of the whole. Samples can be selected randomly or
based on specific features of the data set.
You can learn the basics of predictive analytics and machine learning using a step-by-step tutorial and by building
on samples.
Overview diagram of Azure Machine Learning Studio
capabilities
3/21/2018 • 1 min to read • Edit Online
NOTENOTE
Download the Machine Learning Studio overview diagram
The Microsoft Azure Machine Learning Studio Capabilities Overview diagram gives you a high-level
overview of how you can use Machine Learning Studio to develop a predictive analytics model and operationalize
it in the Azure cloud.
Azure Machine Learning Studio has available a large number of machine learning algorithms, along with modules
that help with data input, output, preparation, and visualization. Using these components you can develop a
predictive analytics experiment, iterate on it, and use it to train your model. Then with one click you can
operationalize your model in the Azure cloud so that it can be used to score new data.
This diagram demonstrates how all those pieces fit together.
See Machine learning algorithm cheat sheet for Microsoft Azure Machine Learning Studio for additional help in navigating
through and choosing the machine learning algorithms available in Machine Learning Studio.
Download the Microsoft Azure Machine Learning Studio Capabilities Overview diagram and get a high-
level view of the capabilities of Machine Learning Studio. To keep it nearby, you can print the diagram in tabloid
size (11 x 17 in.).
Download the diagram here: Microsoft Azure Machine Learning Studio Capabilities Overview
More help with Machine Learning Studio
NOTENOTE
For an overview of Microsoft Azure Machine Learning, see Introduction to machine learning on Microsoft
Azure
For an overview of Machine Learning Studio, see What is Azure Machine Learning Studio?.
For a detailed discussion of the machine learning algorithms available in Machine Learning Studio, see How to
choose algorithms for Microsoft Azure Machine Learning.
You can try Azure Machine Learning for free. No credit card or Azure subscription is required. Get started now.
Downloadable Infographic: Machine learning basics
with algorithm examples
3/21/2018 • 1 min to read • Edit Online
Popular algorithms in Machine Learning Studio
Download the infographic with algorithm examples
More help with algorithms for beginners and advanced users
Download this easy-to-understand infographic overview of machine learning basics to learn about popular
algorithms used to answer common machine learning questions. Algorithm examples help the machine learning
beginner understand which algorithms to use and what they're used for.
Azure Machine Learning Studio comes with a large library of algorithms for predictive analytics. The infographic
identifies four popular families of algorithms - regression, anomaly detection, clustering, and classification - and
provides links to working examples in the Azure AI Gallery. The Gallery contains example experiments and
tutorials that demonstrate how these algorithms can be applied in many real-world solutions.
Download: Infographic of machine learning basics with links to algorithm examples (PDF)
For a deeper discussion of the different types of machine learning algorithms, how they're used, and how to
choose the right one for your solution, see How to choose algorithms for Microsoft Azure Machine Learning.
For a list by category of all the machine learning algorithms available in Machine Learning Studio, see Initialize
Model in the Machine Learning Studio Algorithm and Module Help.
For a complete alphabetical list of algorithms and modules in Machine Learning Studio, see A-Z list of Machine
Learning Studio modules in Machine Learning Studio Algorithm and Module Help.
NOTENOTE
To download and print a diagram that gives an overview of the capabilities of Machine Learning Studio, see
Overview diagram of Azure Machine Learning Studio capabilities.
For an overview of the Azure AI Gallery and the many community-generated resources available there, see
Share and discover resources in the Azure AI Gallery.
You can try Azure Machine Learning for free. No credit card or Azure subscription is required. Get started now.
Azure Machine Learning frequently asked questions:
Billing, capabilities, limitations, and support
4/25/2018 • 29 min to read • Edit Online
General questions
NOTENOTE
Here are some frequently asked questions (FAQs) and corresponding answers about Azure Machine Learning, a
cloud service for developing predictive models and operationalizing solutions through web services. These FAQs
provide questions about how to use the service, which includes the billing model, capabilities, limitations, and
support.
Have a question you can't find here?
Azure Machine Learning has a forum on MSDN where members of the data science community can ask questions
about Azure Machine Learning. The Azure Machine Learning team monitors the forum. Go to the Azure Machine
Learning Forum to search for answers or to post a new question of your own.
What is Azure Machine Learning?
Azure Machine Learning is a fully managed service that you can use to create, test, operate, and manage predictive
analytic solutions in the cloud. With only a browser, you can sign in, upload data, and immediately start machine-
learning experiments. Drag-and-drop predictive modeling, a large pallet of modules, and a library of starting
templates make common machine-learning tasks simple and quick. For more information, see the Azure Machine
Learning service overview. For an introduction to machine learning that explains key terminology and concepts,
see Introduction to Azure Machine Learning.
You can try Azure Machine Learning for free. No credit card or Azure subscription is required. Get started now.
What is Machine Learning Studio?
Machine Learning Studio is a workbench environment that you access by using a web browser. Machine Learning
Studio hosts a pallet of modules in a visual composition interface that helps you build an end-to-end, data-science
workflow in the form of an experiment.
For more information about Machine Learning Studio, see What is Machine Learning Studio?
What is the Machine Learning API service?
The Machine Learning API service enables you to deploy predictive models, like those that are built into Machine
Learning Studio, as scalable, fault-tolerant, web services. The web services that the Machine Learning API service
creates are REST APIs that provide an interface for communication between external applications and your
predictive analytics models.
For more information, see How to consume an Azure Machine Learning Web service.
Where are my Classic web services listed? Where are my New (Azure Resource Manager-based) web
services listed?
Web services created using the Classic deployment model and web services created using the New Azure
Resource Manager deployment model are listed in the Microsoft Azure Machine Learning Web Services portal.
Azure Machine Learning questions
Machine Learning Studio questions
Import and export data for Machine LearningImport and export data for Machine Learning
How large can the data set be for my modules?How large can the data set be for my modules?
Classic web services are also listed in Machine Learning Studio on the Web services tab.
What are Azure Machine Learning web services?
Machine Learning web services provide an interface between an application and a Machine Learning workflow
scoring model. An external application can use Azure Machine Learning to communicate with a Machine Learning
workflow scoring model in real time. A call to a Machine Learning web service returns prediction results to an
external application. To make a call to a web service, you pass an API key that was created when you deployed the
web service. A Machine Learning web service is based on REST, a popular architecture choice for web
programming projects.
Azure Machine Learning has two types of web services:
Request-Response Service (RRS): A low latency, highly scalable service that provides an interface to the
stateless models created and deployed by using Machine Learning Studio.
Batch Execution Service (BES): An asynchronous service that scores a batch for data records.
There are several ways to consume the REST API and access the web service. For example, you can write an
application in C#, R, or Python by using the sample code that's generated for you when you deployed the web
service.
The sample code is available on:
The Consume page for the web service in the Azure Machine Learning Web Services portal
The API Help Page in the web service dashboard in Machine Learning Studio
You can also use the sample Microsoft Excel workbook that's created for you and is available in the web service
dashboard in Machine Learning Studio.
What are the main updates to Azure Machine Learning?
For the latest updates, see What's new in Azure Machine Learning.
What data sources does Machine Learning support?
You can download data to a Machine Learning Studio experiment in three ways:
Upload a local file as a dataset
Use a module to import data from cloud data services
Import a dataset saved from another experiment
To learn more about supported file formats, see Import training data into Machine Learning Studio.
Modules in Machine Learning Studio support datasets of up to 10 GB of dense numerical data for common use
cases. If a module takes more than one input, the 10 GB value is the total of all input sizes. You can also sample
larger datasets by using queries from Hive or Azure SQL Database, or you can use Learning by Counts
preprocessing before ingestion.
The following types of data can expand to larger datasets during feature normalization and are limited to less than
10 GB:
Sparse
What are the limits for data upload?What are the limits for data upload?
ModulesModules
Data processingData processing
Categorical
Strings
Binary data
The following modules are limited to datasets less than 10 GB:
Recommender modules
Synthetic Minority Oversampling Technique (SMOTE) module
Scripting modules: R, Python, SQL
Modules where the output data size can be larger than input data size, such as Join or Feature Hashing
Cross-validation, Tune Model Hyperparameters, Ordinal Regression, and One-vs-All Multiclass, when the
number of iterations is very large
For datasets that are larger than a couple GBs, upload data to Azure Storage or Azure SQL Database, or use Azure
HDInsight rather than directly uploading from a local file.
Can I read data from Amazon S3?
If you have a small amount of data and want to expose it via an HTTP URL, then you can use the Import Data
module. For larger amounts of data, transfer it to Azure Storage first, and then use the Import Data module to
bring it into your experiment.
Is there a built-in image input capability?
You can learn about image input capability in the Import Images reference.
The algorithm, data source, data format, or data transformation operation that I am looking for isn't in
Azure Machine Learning Studio. What are my options?
You can go to the user feedback forum to see feature requests that we are tracking. Add your vote to a request if a
capability that you're looking for has already been requested. If the capability that you're looking for doesn't exist,
create a new request. You can view the status of your request in this forum, too. We track this list closely and
update the status of feature availability frequently. In addition, you can use the built-in support for R and Python to
create custom transformations when needed.
Can I bring my existing code into Machine Learning Studio?
Yes, you can bring your existing R or Python code into Machine Learning Studio, run it in the same experiment with
Azure Machine Learning learners, and deploy the solution as a web service via Azure Machine Learning. For more
information, see Extend your experiment with R and Execute Python machine learning scripts in Azure Machine
Learning Studio.
Is it possible to use something like PMML to define a model?
No, Predictive Model Markup Language (PMML) is not supported. You can use custom R and Python code to
define a module.
How many modules can I execute in parallel in my experiment?
You can execute up to four modules in parallel in an experiment.
Is there an ability to visualize data (beyond R visualizations) interactively within the experiment?
Click the output of a module to visualize the data and get statistics.
When previewing results or data in a browser, the number of rows and columns is limited. Why?
AlgorithmsAlgorithms
R moduleR module
Python modulePython module
Because large amounts of data might be sent to a browser, data size is limited to prevent slowing down Machine
Learning Studio. To visualize all the data/results, it's better to download the data and use Excel or another tool.
What existing algorithms are supported in Machine Learning Studio?
Machine Learning Studio provides state-of-the-art algorithms, such as Scalable Boosted Decision trees, Bayesian
Recommendation systems, Deep Neural Networks, and Decision Jungles developed at Microsoft Research.
Scalable open-source machine learning packages, like Vowpal Wabbit, are also included. Machine Learning Studio
supports machine learning algorithms for multiclass and binary classification, regression, and clustering. See the
complete list of Machine Learning Modules.
Do you automatically suggest the right Machine Learning algorithm to use for my data?
No, but Machine Learning Studio has various ways to compare the results of each algorithm to determine the right
one for your problem.
Do you have any guidelines on picking one algorithm over another for the provided algorithms?
See How to choose an algorithm.
Are the provided algorithms written in R or Python?
No, these algorithms are mostly written in compiled languages to provide better performance.
Are any details of the algorithms provided?
The documentation provides some information about the algorithms and parameters for tuning are described to
optimize the algorithm for your use.
Is there any support for online learning?
No, currently only programmatic retraining is supported.
Can I visualize the layers of a Neural Net Model by using the built-in module?
No.
Can I create my own modules in C# or some other language?
Currently, you can only use R to create new custom modules.
What R packages are available in Machine Learning Studio?
Machine Learning Studio supports more than 400 CRAN R packages today, and here is the current list of all
included packages. Also, see Extend your experiment with R to learn how to retrieve this list yourself. If the package
that you want is not in this list, provide the name of the package at the user feedback forum.
Is it possible to build a custom R module?
Yes, see Author custom R modules in Azure Machine Learning for more information.
Is there a REPL environment for R?
No, there is no Read-Eval-Print-Loop (REPL) environment for R in the studio.
Is it possible to build a custom Python module?
Not currently, but you can use one or more Execute Python Script modules to get the same result.
Web service
RetrainRetrain
CreateCreate
UseUse
Is there a REPL environment for Python?
You can use the Jupyter Notebooks in Machine Learning Studio. For more information, see Introducing Jupyter
Notebooks in Azure Machine Learning Studio.
How do I retrain Azure Machine Learning models programmatically?
Use the retraining APIs. For more information, see Retrain Machine Learning models programmatically. Sample
code is also available in the Microsoft Azure Machine Learning Retraining Demo.
Can I deploy the model locally or in an application that doesn't have an Internet connection?
No.
Is there a baseline latency that is expected for all web services?
See the Azure subscription limits.
When would I want to run my predictive model as a Batch Execution service versus a Request Response
service?
The Request Response service (RRS) is a low-latency, high-scale web service that is used to provide an interface to
stateless models that are created and deployed from the experimentation environment. The Batch Execution
service (BES) is a service that asynchronously scores a batch of data records. The input for BES is like data input
that RRS uses. The main difference is that BES reads a block of records from a variety of sources, such as Azure
Blob storage, Azure Table storage, Azure SQL Database, HDInsight (hive query), and HTTP sources. For more
information, see How to consume an Azure Machine Learning Web service.
How do I update the model for the deployed web service?
To update a predictive model for an already deployed service, modify and rerun the experiment that you used to
author and save the trained model. After you have a new version of the trained model available, Machine Learning
Studio asks you if you want to update your web service. For details about how to update a deployed web service,
see Deploy a Machine Learning web service.
You can also use the Retraining APIs. For more information, see Retrain Machine Learning models
programmatically. Sample code is also available in the Microsoft Azure Machine Learning Retraining Demo.
How do I monitor my web service deployed in production?
After you deploy a predictive model, you can monitor it from the Azure Machine Learning Web Services portal.
Each deployed service has its own dashboard where you can see monitoring information for that service. For more
information about how to manage your deployed web services, see Manage a Web service using the Azure
Machine Learning Web Services portal and Manage an Azure Machine Learning workspace.
Is there a place where I can see the output of my RRS/BES?
For RRS, the web service response is typically where you see the result. You can also write it to Azure Blob storage.
For BES, the output is written to a blob by default. You can also write the output to a database or table by using the
Export Data module.
Can I create web services only from models that were created in Machine Learning Studio?
No, you can also create web services directly by using Jupyter Notebooks and RStudio.
Scalability
Security and availability
Where can I find information about error codes?
See Machine Learning Module Error Codes for a list of error codes and descriptions.
What is the scalability of the web service?
Currently, the default endpoint is provisioned with 20 concurrent RRS requests per endpoint. You can scale this to
200 concurrent requests per endpoint, and you can scale each web service to 10,000 endpoints per web service as
described in Scaling a Web Service. For BES, each endpoint can process 40 requests at a time, and additional
requests beyond 40 requests are queued. These queued requests run automatically as the queue drains.
Are R jobs spread across nodes?
No.
How much data can I use for training?
Modules in Machine Learning Studio support datasets of up to 10 GB of dense numerical data for common use
cases. If a module takes more than one input, the total size for all inputs is 10 GB. You can also sample larger
datasets via Hive queries, via Azure SQL Database queries, or by preprocessing with Learning with Counts
modules before ingestion.
The following types of data can expand to larger datasets during feature normalization and are limited to less than
10 GB:
Sparse
Categorical
Strings
Binary data
The following modules are limited to datasets less than 10 GB:
Recommender modules
Synthetic Minority Oversampling Technique (SMOTE) module
Scripting modules: R, Python, SQL
Modules where the output data size can be larger than input data size, such as Join or Feature Hashing
Cross-Validate, Tune Model Hyperparameters, Ordinal Regression, and One-vs-All Multiclass, when number of
iterations is very large
For datasets that are larger than a few GBs, upload data to Azure Storage or Azure SQL Database, or use
HDInsight rather than directly uploading from a local file.
Are there any vector size limitations?
Rows and columns are each limited to the .NET limitation of Max Int: 2,147,483,647.
Can I adjust the size of the virtual machine that runs the web service?
No.
Who can access the http endpoint for the web service by default? How do I restrict access to the
endpoint?
After a web service is deployed, a default endpoint is created for that service. The default endpoint can be called by
Support and training
Billing questions
using its API key. You can add more endpoints with their own keys from the Web Services portal or
programmatically by using the Web Service Management APIs. Access keys are needed to make calls to the web
service. For more information, see How to consume an Azure Machine Learning Web service.
What happens if my Azure storage account can't be found?
Machine Learning Studio relies on a user-supplied Azure storage account to save intermediary data when it
executes the workflow. This storage account is provided to Machine Learning Studio when a workspace is created.
After the workspace is created, if the storage account is deleted and can no longer be found, the workspace will
stop functioning, and all experiments in that workspace will fail.
If you accidentally deleted the storage account, recreate the storage account with the same name in the same
region as the deleted storage account. After that, resync the access key.
What happens if my storage account access key is out of sync?
Machine Learning Studio relies on a user-supplied Azure storage account to store intermediary data when it
executes the workflow. This storage account is provided to Machine Learning Studio when a workspace is created,
and the access keys are associated with that workspace. If the access keys are changed after the workspace is
created, the workspace can no longer access the storage account. It will stop functioning and all experiments in that
workspace will fail.
If you changed storage account access keys, resync the access keys in the workspace by using the Azure portal.
Where can I get training for Azure Machine Learning?
The Azure Machine Learning Documentation Center hosts video tutorials and how-to guides. These step-by-step
guides introduce the services and explain the data science life cycle of importing data, cleaning data, building
predictive models, and deploying them in production by using Azure Machine Learning.
We add new material to the Machine Learning Center on an ongoing basis. You can submit requests for additional
learning material on Machine Learning Center at the user feedback forum.
You can also find training at Microsoft Virtual Academy.
How do I get support for Azure Machine Learning?
To get technical support for Azure Machine Learning, go to Azure Support, and select Machine Learning.
Azure Machine Learning also has a community forum on MSDN where you can ask questions about Azure
Machine Learning. The Azure Machine Learning team monitors the forum. Go to Azure Forum.
How does Machine Learning billing work?
Azure Machine Learning has two components: Machine Learning Studio and Machine Learning web services.
While you are evaluating Machine Learning Studio, you can use the Free billing tier. The Free tier also lets you
deploy a Classic web service that has limited capacity.
If you decide that Azure Machine Learning meets your needs, you can sign up for the Standard tier. To sign up, you
must have a Microsoft Azure subscription.
In the Standard tier, you are billed monthly for each workspace that you define in Machine Learning Studio. When
you run an experiment in the studio, you are billed for compute resources when you are running an experiment.
When you deploy a Classic web service, transactions and compute hours are billed on the Pay As You Go basis.
NOTENOTE
New (Resource Manager-based) web services introduce billing plans that allow for more predictability in costs.
Tiered pricing offers discounted rates to customers who need a large amount of capacity.
When you create a plan, you commit to a fixed cost that comes with an included quantity of API compute hours
and API transactions. If you need more included quantities, you can add instances to your plan. If you need a lot
more included quantities, you can choose a higher tier plan that provides considerably more included quantities
and a better discounted rate.
After the included quantities in existing instances are used up, additional usage is charged at the overage rate that's
associated with the billing plan tier.
Included quantities are reallocated every 30 days, and unused included quantities do not roll over to the next period.
For additional billing and pricing information, see Machine Learning Pricing.
Does Machine Learning have a free trial?
Azure Machine Learning has a free subscription option that's explained in Machine Learning Pricing. Machine
Learning Studio has an eight-hour quick evaluation trial that's available when you sign in to Machine Learning
Studio.
In addition, when you sign up for an Azure free trial, you can try any Azure services for a month. To learn more
about the Azure free trial, visit Azure free trial FAQ.
What is a transaction?
A transaction represents an API call that Azure Machine Learning responds to. Transactions from Request-
Response Service (RRS) and Batch Execution Service (BES) calls are aggregated and charged against your billing
plan.
Can I use the included transaction quantities in a plan for both RRS and BES transactions?
Yes, your transactions from your RRS and BES are aggregated and charged against your billing plan.
What is an API compute hour?
An API compute hour is the billing unit for the time that API calls take to run by using Machine Learning compute
resources. All your calls are aggregated for billing purposes.
How long does a typical production API call take?
Production API call times can vary significantly, generally ranging from hundreds of milliseconds to a few seconds.
Some API calls might require minutes depending on the complexity of the data processing and machine-learning
model. The best way to estimate production API call times is to benchmark a model on the Machine Learning
service.
What is a Studio compute hour?
A Studio compute hour is the billing unit for the aggregate time that your experiments use compute resources in
studio.
In New (Azure Resource Manager-based) web services, what is the Dev/Test tier meant for?
Resource Manager-based web services provide multiple tiers that you can use to provision your billing plan. The
Dev/Test pricing tier provides limited, included quantities that allow you to test your experiment as a web service
without incurring costs. You have the opportunity to see how it works.
Are there separate storage charges?
Management of New (Resource Manager-based) web servicesManagement of New (Resource Manager-based) web services
NOTENOTE
NOTENOTE
NOTENOTE
Sign up for New (Resource Manager-based) web services plansSign up for New (Resource Manager-based) web services plans
The Machine Learning Free tier does not require or allow separate storage. The Machine Learning Standard tier
requires users to have an Azure storage account. Azure Storage is billed separately.
Does Machine Learning support high availability?
Yes. For details, see Machine Learning Pricing for a description of the service level agreement (SL A).
What specific kind of compute resources will my production API calls be run on?
The Machine Learning service is a multitenant service. Actual compute resources that are used on the back end
vary and are optimized for performance and predictability.
What happens if I delete my plan?
The plan is removed from your subscription, and you are billed for prorated usage.
You cannot delete a plan that a web service is using. To delete the plan, you must either assign a new plan to the web service
or delete the web service.
What is a plan instance?
A plan instance is a unit of included quantities that you can add to your billing plan. When you select a billing tier
for your billing plan, it comes with one instance. If you need more included quantities, you can add instances of the
selected billing tier to your plan.
How many plan instances can I add?
You can have one instance of the Dev/Test pricing tier in a subscription.
For Standard S1, Standard S2, and Standard S3 tiers, you can add as many as necessary.
Depending on your anticipated usage, it might be more cost effective to upgrade to a tier that has more included quantities
rather than add instances to the current tier.
What happens when I change plan tiers (upgrade / downgrade)?
The old plan is deleted and the current usage is billed on a prorated basis. A new plan with the full included
quantities of the upgraded/downgraded tier is created for the rest of the period.
Included quantities are allocated per period, and unused quantities do not roll over.
What happens when I increase the instances in a plan?
Quantities are included on a prorated basis and may take 24 hours to be effective.
What happens when I delete an instance of a plan?
The instance is removed from your subscription, and you are billed for prorated usage.
How do I sign up for a plan?
New web services: OveragesNew web services: Overages
You have two ways to create billing plans.
When you first deploy a Resource Manager-based web service, you can choose an existing plan or create a new
plan.
Plans that you create in this manner are in your default region, and your web service will be deployed to that
region.
If you want to deploy services to regions other than your default region, you may want to define your billing plans
before you deploy your service.
In that case, you can sign in to the Azure Machine Learning Web Services portal, and go to the Plans page. From
there, you can add plans, delete plans, and modify existing plans.
Which plan should I choose to start off with?
We recommend you that you start with the Standard S1 tier and monitor your service for usage. If you find that
you are using your included quantities rapidly, you can add instances or move to a higher tier and get better
discounted rates. You can adjust your billing plan as needed throughout your billing cycle.
Which regions are the new plans available in?
The new billing plans are available in the three production regions in which we support the new web services:
South Central US
West Europe
South East Asia
I have web services in multiple regions. Do I need a plan for every region?
Yes. Plan pricing varies by region. When you deploy a web service to another region, you need to assign it a plan
that is specific to that region. For more information, see Products available by region.
How do I check if I exceeded my web service usage?
You can view the usage on all your plans on the Plans page in the Azure Machine Learning Web Services portal.
Sign in to the portal, and then click the Plans menu option.
In the Transactions and Compute columns of the table, you can see the included quantities of the plan and the
percentage used.
What happens when I use up the include quantities in the Dev/Test pricing tier?
Services that have a Dev/Test pricing tier assigned to them are stopped until the next period or until you move
them to a paid tier.
For Classic web services and overages of New (Resource Manager-based) web services, how are prices
calculated for Request Response (RRS) and Batch (BES) workloads?
For an RRS workload, you are charged for every API transaction call that you make and for the compute time
that's associated with those requests. Your RRS production API transaction costs are calculated as the total number
of API calls that you make multiplied by the price per 1,000 transactions (prorated by individual transaction). Your
RRS API production API compute hour costs are calculated as the amount of time required for each API call to run,
multiplied by the total number of API transactions, multiplied by the price per production API compute hour.
For example, for Standard S1 overage, 1,000,000 API transactions that take 0.72 seconds each to run would result
in (1,000,000 * $0.50/1K API transactions) in $500 in production API transaction costs and (1,000,000 * 0.72 sec *
$2/hr) $400 in production API compute hours, for a total of $900.
Azure Machine Learning Classic web servicesAzure Machine Learning Classic web services
Azure Machine Learning Free and Standard tierAzure Machine Learning Free and Standard tier
For a BES workload, you are charged in the same manner. However, the API transaction costs represent the
number of batch jobs that you submit, and the compute costs represent the compute time that's associated with
those batch jobs. Your BES production API transaction costs are calculated as the total number of jobs submitted
multiplied by the price per 1,000 transactions (prorated by individual transaction). Your BES API production API
compute hour costs are calculated as the amount of time required for each row in your job to run multiplied by the
total number of rows in your job multiplied by the total number of jobs multiplied by the price per production API
compute hour. When you use the Machine Learning calculator, the transaction meter represents the number of jobs
that you plan to submit, and the time-per-transaction field represents the combined time that's needed for all rows
in each job to run.
For example, assume Standard S1 overage, and you submit 100 jobs per day that each consist of 500 rows that
take 0.72 seconds each. Your monthly overage costs would be (100 jobs per day = 3,100 jobs/mo * $0.50/1K API
transactions) $1.55 in production API transaction costs and (500 rows * 0.72 sec * 3,100 Jobs * $2/hr) $620 in
production API compute hours, for a total of $621.55.
Is Pay As You Go still available?
Yes, Classic web services are still available in Azure Machine Learning.
What is included in the Azure Machine Learning Free tier?
The Azure Machine Learning Free tier is intended to provide an in-depth introduction to the Azure Machine
Learning Studio. All you need is a Microsoft account to sign up. The Free tier includes free access to one Azure
Machine Learning Studio workspace per Microsoft account. In this tier, you can use up to 10 GB of storage and
operationalize models as staging APIs. Free tier workloads are not covered by an SL A and are intended for
development and personal use only.
Free tier workspaces have the following limitations:
Workloads can't access data by connecting to an on-premises server that runs SQL Server.
You cannot deploy New Resource Manager base web services.
What is included in the Azure Machine Learning Standard tier and plans?
The Azure Machine Learning Standard tier is a paid production version of Azure Machine Learning Studio. The
Azure Machine Learning Studio monthly fee is billed on a per workspace per month basis and prorated for partial
months. Azure Machine Learning Studio experiment hours are billed per compute hour for active experimentation.
Billing is prorated for partial hours.
The Azure Machine Learning API service is billed depending on whether it's a Classic web service or a New
(Resource Manager-based) web service.
The following charges are aggregated per workspace for your subscription.
Machine Learning Workspace Subscription: The Machine Learning workspace subscription is a monthly fee that
provides access to a Machine Learning Studio workspace. The subscription is required to run experiments in the
studio and to utilize the production APIs.
Studio Experiment hours: This meter aggregates all compute charges that are accrued by running experiments
in Machine Learning Studio and running production API calls in the staging environment.
Access data by connecting to an on-premises server that runs SQL Server in your models for your training and
scoring.
For Classic web services:
Production API Compute Hours: This meter includes compute charges that are accrued by web services
running in production.
Production API Transactions (in 1000s): This meter includes charges that are accrued per call to your
production web service.
Apart from the preceding charges, in the case of Resource Manager-based web service, charges are aggregated to
the selected plan:
Standard S1/S2/S3 API Plan (Units): This meter represents the type of instance that's selected for Resource
Manager-based web services.
Standard S1/S2/S3 Overage API Compute Hours: This meter includes compute charges that are accrued by
Resource Manager-based web services that run in production after the included quantities in existing instances
are used up. The additional usage is charged at the overate rate that's associated with S1/S2/S3 plan tier.
Standard S1/S2/S3 Overage API Transactions (in 1,000s): This meter includes charges that are accrued per call
to your production Resource Manager-based web service after the included quantities in existing instances are
used up. The additional usage is charged at the overate rate associated with S1/S2/S3 plan tier.
Included Quantity API Compute Hours: With Resource Manager-based web services, this meter represents the
included quantity of API compute hours.
Included Quantity API Transactions (in 1,000s): With Resource Manager-based web services, this meter
represents the included quantity of API transactions.
How do I sign up for Azure Machine Learning Free tier?
All you need is a Microsoft account. Go to Azure Machine Learning home, and then click Start Now. Sign in with
your Microsoft account and a workspace in Free tier is created for you. You can start to explore and create Machine
Learning experiments right away.
How do I sign up for Azure Machine Learning Standard tier?
You must first have access to an Azure subscription to create a Standard Machine Learning workspace. You can
sign up for a 30-day free trial Azure subscription and later upgrade to a paid Azure subscription, or you can
purchase a paid Azure subscription outright. You can then create a Machine Learning workspace from the
Microsoft Azure portal after you gain access to the subscription. View the step-by-step instructions.
Alternatively, you can be invited by a Standard Machine Learning workspace owner to access the owner's
workspace.
Can I specify my own Azure Blob storage account to use with the Free tier?
No, the Standard tier is equivalent to the version of the Machine Learning service that was available before the
tiers were introduced.
Can I deploy my machine learning models as APIs in the Free tier?
Yes, you can operationalize machine learning models to staging API services as part of the Free tier. To put the
staging API service into production and get a production endpoint for the operationalized service, you must use
the Standard tier.
What is the difference between Azure free trial and Azure Machine Learning Free tier?
The Microsoft Azure free trial offers credits that you can apply to any Azure service for one month. The Azure
Machine Learning Free tier offers continuous access specifically to Azure Machine Learning for non-production
workloads.
How do I move an experiment from the Free tier to the Standard tier?
To copy your experiments from the Free tier to the Standard tier:
1. Sign in to Azure Machine Learning Studio, and make sure that you can see both the Free workspace and the
Standard workspace in the workspace selector in the top navigation bar.
Studio workspaceStudio workspace
Guest AccessGuest Access
2. Switch to Free workspace if you are in the Standard workspace.
3. In the experiment list view, select an experiment that you'd like to copy, and then click the Copy command
button.
4. Select the Standard workspace from the dialog box that opens, and then click the Copy button. All the
associated datasets, trained model, etc. are copied together with the experiment into the Standard workspace.
5. You need to rerun the experiment and republish your web service in the Standard workspace.
Will I see different bills for different workspaces?
Workspace charges are broken out separately for each applicable meter on a single bill.
What specific kind of compute resources will my experiments be run on?
The Machine Learning service is a multitenant service. Actual compute resources that are used on the back end
vary and are optimized for performance and predictability.
What is Guest Access to Azure Machine Learning Studio?
Guest Access is a restricted trial experience. You can create and run experiments in Azure Machine Learning Studio
at no cost and without authentication. Guest sessions are non-persistent (cannot be saved) and limited to eight
hours. Other limitations include lack of support for R and Python, lack of staging APIs, and restricted dataset size
and storage capacity. By comparison, users who choose to sign in with a Microsoft account have full access to the
Free tier of Machine Learning Studio that's described previously, which includes a persistent workspace and more
comprehensive capabilities. To choose your free Machine Learning experience, click Get started on
https://studio.azureml.net, and then select Guess Access or sign in with a Microsoft account.
What's New in Azure Machine Learning
3/21/2018 • 1 min to read • Edit Online
The March 2017 release of Microsoft Azure Machine Learning updates provides the following feature:The March 2017 release of Microsoft Azure Machine Learning updates provides the following feature:
The August 2016 release of Microsoft Azure Machine Learning updates provide the following features:The August 2016 release of Microsoft Azure Machine Learning updates provide the following features:
The July 2016 release of Microsoft Azure Machine Learning updates provide the following features:The July 2016 release of Microsoft Azure Machine Learning updates provide the following features:
Dedicated Capacity for Azure Machine Learning BES Jobs
Machine Learning Batch Pool processing uses the Azure Batch service to provide customer-managed scale
for the Azure Machine Learning Batch Execution Service. Batch Pool processing allows you to create Azure
Batch pools on which you can submit batch jobs and have them execute in a predictable manner.
For more information, see Azure Batch service for Machine Learning jobs.
Classic Web services can now be managed in the new Microsoft Azure Machine Learning Web Services portal
that provides one place to manage all aspects of your Web service.
Which provides web service usage statistics.
Simplifies testing of Azure Machine Learning Remote-Request calls using sample data.
Provides a new Batch Execution Service test page with sample data and job submission history.
Provides easier endpoint management.
Web services are now managed as Azure resources managed through Azure Resource Manager interfaces,
allowing for the following enhancements:
Incorporates a new subscription-based, multi-region web service deployment model using Resource Manager
based APIs leveraging the Resource Manager Resource Provider for Web Services.
Introduces new pricing plans and plan management capabilities using the new Resource Manager RP for
Billing.
Provides web service usage statistics.
Simplifies testing of Azure Machine Learning Remote-Request calls using sample data.
Provides a new Batch Execution Service test page with sample data and job submission history.
There are new REST APIs to deploy and manage your Resource Manager based Web services.
There is a new Microsoft Azure Machine Learning Web Services portal that provides one place to
manage all aspects of your Web service.
You can now deploy your web service to multiple regions without needing to create a subscription in
each region.
In addition, the Machine Learning Studio has been updated to allow you to deploy to the new Web service model
or continue to deploy to the classic Web service model.
NOTENOTE
NOTENOTE
How does Machine Learning Studio help?
If you've never used Azure Machine Learning Studio before, this tutorial is for you.
In this tutorial, we'll walk through how to use Studio for the first time to create a machine learning experiment.
The experiment will test an analytical model that predicts the price of an automobile based on different variables
such as make and technical specifications.
This tutorial shows you the basics of how to drag-and-drop modules onto your experiment, connect them together, run
the experiment, and look at the results. We're not going to discuss the general topic of machine learning or how to select
and use the 100+ built-in algorithms and data manipulation modules included in Studio.
If you're new to machine learning, the video series Data Science for Beginners might be a good place to start. This video
series is a great introduction to machine learning using everyday language and concepts.
If you're familiar with machine learning, but you're looking for more general information about Machine Learning Studio,
and the machine learning algorithms it contains, here are some good resources:
What is Machine Learning Studio? - This is a high-level overview of Studio.
Machine learning basics with algorithm examples - This infographic is useful if you want to learn more about the
different types of machine learning algorithms included with Machine Learning Studio.
Machine Learning Guide - This guide covers similar information as the infographic above, but in an interactive format.
Machine learning algorithm cheat sheet and How to choose algorithms for Microsoft Azure Machine Learning - This
downloadable poster and accompanying article discuss the Studio algorithms in depth.
Machine Learning Studio: Algorithm and Module Help - This is the complete reference for all Studio modules, including
machine learning algorithms,
You can try Azure Machine Learning for free. No credit card or Azure subscription is required. Get started now.
Machine Learning Studio makes it easy to set up an experiment using drag-and-drop modules preprogrammed
with predictive modeling techniques.
Using an interactive, visual workspace, you drag-and-drop datasets and analysis modules onto an interactive
canvas. You connect them together to form an experiment that you run in Machine Learning Studio. You create
a model, train the model, and score and test the model.
You can iterate on your model design, editing the experiment and running it until it gives you the results you're
looking for. When your model is ready, you can publish it as a web service so that others can send it new data
and get predictions in return.
Open Machine Learning Studio
Five steps to create an experiment
TIPTIP
Step 1: Get data
To get started with Studio, go to https://studio.azureml.net. If you’ve signed into Machine Learning Studio before,
click Sign In. Otherwise, click Sign up here and choose between free and paid options.
Sign in to Machine Learning Studio
In this machine learning tutorial, you'll follow five basic steps to build an experiment in Machine Learning Studio
to create, train, and score your model:
Create a model
Train the model
Score and test the model
Step 1: Get data
Step 2: Prepare the data
Step 3: Define features
Step 4: Choose and apply a learning algorithm
Step 5: Predict new automobile prices
You can find a working copy of the following experiment in the Azure AI Gallery. Go to Your first data science
experiment - Automobile price prediction and click Open in Studio to download a copy of the experiment into your
Machine Learning Studio workspace.
The first thing you need to perform machine learning is data. There are several sample datasets included with
Machine Learning Studio that you can use, or you can import data from many sources. For this example, we'll
use the sample dataset, Automobile price data (Raw), that's included in your workspace. This dataset includes
entries for various individual automobiles, including information such as make, model, technical specifications,
and price.
Here's how to get the dataset into your experiment.
1. Create a new experiment by clicking +NEW at the bottom of the Machine Learning Studio window, select
EXPERIMENT, and then select Blank Experiment.
2. The experiment is given a default name that you can see at the top of the canvas. Select this text and
rename it to something meaningful, for example, Automobile price prediction. The name doesn't need
to be unique.
3. To the left of the experiment canvas is a palette of datasets and modules. Type automobile in the Search
box at the top of this palette to find the dataset labeled Automobile price data (Raw). Drag this dataset
to the experiment canvas.
Find the automobile dataset and drag it onto the experiment canvas
To see what this data looks like, click the output port at the bottom of the automobile dataset, and then select
Visualize.
Click the output port and select "Visualize"
TIPTIP
Step 2: Prepare the data
TIPTIP
Datasets and modules have input and output ports represented by small circles - input ports at the top, output ports at
the bottom. To create a flow of data through your experiment, you'll connect an output port of one module to an input
port of another. At any time, you can click the output port of a dataset or module to see what the data looks like at that
point in the data flow.
In this sample dataset, each instance of an automobile appears as a row, and the variables associated with each
automobile appear as columns. Given the variables for a specific automobile, we're going to try to predict the
price in far-right column (column 26, titled "price").
View the automobile data in the data visualization window
Close the visualization window by clicking the "x" in the upper-right corner.
A dataset usually requires some preprocessing before it can be analyzed. For example, you might have noticed
the missing values present in the columns of various rows. These missing values need to be cleaned so the
model can analyze the data correctly. In our case, we'll remove any rows that have missing values. Also, the
normalized-losses column has a large proportion of missing values, so we'll exclude that column from the
model altogether.
Cleaning the missing values from input data is a prerequisite for using most of the modules.
First we add a module that removes the normalized-losses column completely, and then we add another
module that removes any row that has missing data.
1. Type select columns in the Search box at the top of the module palette to find the Select Columns in
Dataset module, then drag it to the experiment canvas. This module allows us to select which columns of
data we want to include or exclude in the model.
2. Connect the output port of the Automobile price data (Raw) dataset to the input port of the Select
Columns in Dataset module.
Add the "Select Columns in Dataset" module to the experiment canvas and connect it
3. Click the Select Columns in Dataset module and click Launch column selector in the Properties pane.
On the left, click With rules
Under Begin With, click All columns. This directs Select Columns in Dataset to pass through all the
columns (except those columns we're about to exclude).
From the drop-downs, select Exclude and column names, and then click inside the text box. A list of
columns is displayed. Select normalized-losses, and it's added to the text box.
Click the check mark (OK) button to close the column selector (on the lower-right).
Launch the column selector and exclude the "normalized-losses" column
Now the properties pane for Select Columns in Dataset indicates that it will pass through all
columns from the dataset except normalized-losses.
The properties pane shows that the "normalized-losses" column is excluded
TIPTIP
You can add a comment to a module by double-clicking the module and entering text. This can help you
see at a glance what the module is doing in your experiment. In this case double-click the Select Columns in
Dataset module and type the comment "Exclude normalized losses."
Double-click a module to add a comment
4. Drag the Clean Missing Data module to the experiment canvas and connect it to the Select Columns in
Dataset module. In the Properties pane, select Remove entire row under Cleaning mode. This directs
Clean Missing Data to clean the data by removing rows that have any missing values. Double-click the
module and type the comment "Remove missing value rows."
Set the cleaning mode to "Remove entire row" for the "Clean Missing Data" module
5. Run the experiment by clicking RUN at the bottom of the page.
When the experiment has finished running, all the modules have a green check mark to indicate that they
finished successfully. Notice also the Finished running status in the upper-right corner.
TIPTIP
Step 3: Define features
make, body-style, wheel-base, engine-size, horsepower, peak-rpm, highway-mpg, price
After running it, the experiment should look something like this
Why did we run the experiment now? By running the experiment, the column definitions for our data pass from the
dataset, through the Select Columns in Dataset module, and through the Clean Missing Data module. This means that any
modules we connect to Clean Missing Data will also have this same information.
All we have done in the experiment up to this point is clean the data. If you want to view the cleaned dataset, click
the left output port of the Clean Missing Data module and select Visualize. Notice that the normalized-losses
column is no longer included, and there are no missing values.
Now that the data is clean, we're ready to specify what features we're going to use in the predictive model.
In machine learning, features are individual measurable properties of something you’re interested in. In our
dataset, each row represents one automobile, and each column is a feature of that automobile.
Finding a good set of features for creating a predictive model requires experimentation and knowledge about the
problem you want to solve. Some features are better for predicting the target than others. Also, some features
have a strong correlation with other features and can be removed. For example, city-mpg and highway-mpg are
closely related so we can keep one and remove the other without significantly affecting the prediction.
Let's build a model that uses a subset of the features in our dataset. You can come back later and select different
features, run the experiment again, and see if you get better results. But to start, let's try the following features:
1. Drag another Select Columns in Dataset module to the experiment canvas. Connect the left output port of
the Clean Missing Data module to the input of the Select Columns in Dataset module.
Step 4: Choose and apply a learning algorithm
Connect the "Select Columns in Dataset" module to the "Clean Missing Data" module
2. Double-click the module and type "Select features for prediction."
3. Click Launch column selector in the Properties pane.
4. Click With rules.
5. Under Begin With, click No columns. In the filter row, select Include and column names and select
our list of column names in the text box. This directs the module to not pass through any columns
(features) except the ones that we specify.
6. Click the check mark (OK) button.
Select the columns (features) to include in the prediction
This produces a filtered dataset containing only the features we want to pass to the learning algorithm we'll use
in the next step. Later, you can return and try again with a different selection of features.
Now that the data is ready, constructing a predictive model consists of training and testing. We'll use our data to
train the model, and then we'll test the model to see how closely it's able to predict prices.
Classification and regression are two types of supervised machine learning algorithms. Classification predicts an
answer from a defined set of categories, such as a color (red, blue, or green). Regression is used to predict a
number.
Because we want to predict price, which is a number, we'll use a regression algorithm. For this example, we'll use
a simple linear regression model.
TIPTIP
If you want to learn more about different types of machine learning algorithms and when to use them, you might view the
first video in the Data Science for Beginners series, The five questions data science answers. You might also look at the
infographic Machine learning basics with algorithm examples, or check out the Machine learning algorithm cheat sheet.
We train the model by giving it a set of data that includes the price. The model scans the data and look for
correlations between an automobile's features and its price. Then we'll test the model - we'll give it a set of
features for automobiles we're familiar with and see how close the model comes to predicting the known price.
We'll use our data for both training the model and testing it by splitting the data into separate training and
testing datasets.
TIPTIP
1. Select and drag the Split Data module to the experiment canvas and connect it to the last Select Columns
in Dataset module.
2. Click the Split Data module to select it. Find the Fraction of rows in the first output dataset (in the
Properties pane to the right of the canvas) and set it to 0.75. This way, we'll use 75 percent of the data to
train the model, and hold back 25 percent for testing (later, you can experiment with using different
percentages).
Set the split fraction of the "Split Data" module to 0.75
By changing the Random seed parameter, you can produce different random samples for training and testing.
This parameter controls the seeding of the pseudo-random number generator.
3. Run the experiment. When the experiment is run, the Select Columns in Dataset and Split Data modules
pass column definitions to the modules we'll be adding next.
4. To select the learning algorithm, expand the Machine Learning category in the module palette to the left
of the canvas, and then expand Initialize Model. This displays several categories of modules that can be
used to initialize machine learning algorithms. For this experiment, select the Linear Regression module
under the Regression category, and drag it to the experiment canvas. (You can also find the module by
typing "linear regression" in the palette Search box.)
5. Find and drag the Train Model module to the experiment canvas. Connect the output of the Linear
Regression module to the left input of the Train Model module, and connect the training data output (left
port) of the Split Data module to the right input of the Train Model module.
Connect the "Train Model" module to both the "Linear Regression" and "Split Data" modules
6. Click the Train Model module, click Launch column selector in the Properties pane, and then select the
price column. This is the value that our model is going to predict.
You select the price column in the column selector by moving it from the Available columns list to the
Selected columns list.
Select the price column for the "Train Model" module
7. Run the experiment.
We now have a trained regression model that can be used to score new automobile data to make price
predictions.
Step 5: Predict new automobile prices
After running, the experiment should now look something like this
Now that we've trained the model using 75 percent of our data, we can use it to score the other 25 percent of the
data to see how well our model functions.
1. Find and drag the Score Model module to the experiment canvas. Connect the output of the Train Model
module to the left input port of Score Model. Connect the test data output (right port) of the Split Data
module to the right input port of Score Model.
Connect the "Score Model" module to both the "Train Model" and "Split Data" modules
TIPTIP
2. Run the experiment and view the output from the Score Model module (click the output port of Score
Model and select Visualize). The output shows the predicted values for price and the known values from
the test data.
Output of the "Score Model" module
3. Finally, we test the quality of the results. Select and drag the Evaluate Model module to the experiment
canvas, and connect the output of the Score Model module to the left input of Evaluate Model.
There are two input ports on the Evaluate Model module because it can be used to compare two models side by
side. Later, you can add another algorithm to the experiment and use Evaluate Model to see which one gives better
results.
4. Run the experiment.
To view the output from the Evaluate Model module, click the output port, and then select Visualize.
Final experiment
Evaluation results for the experiment
The following statistics are shown for our model:
Mean Absolute Error (MAE): The average of absolute errors (an error is the difference between the
predicted value and the actual value).
Root Mean Squared Error (RMSE): The square root of the average of squared errors of predictions made
on the test dataset.
Relative Absolute Error: The average of absolute errors relative to the absolute difference between actual
values and the average of all actual values.
Relative Squared Error: The average of squared errors relative to the squared difference between the actual
values and the average of all actual values.
Coefficient of Determination: Also known as the R squared value, this is a statistical metric indicating
how well a model fits the data.
For each of the error statistics, smaller is better. A smaller value indicates that the predictions more closely match
the actual values. For Coefficient of Determination, the closer its value is to one (1.0), the better the
predictions.
The final experiment should look something like this:
Next steps
The final experiment
Now that you've completed the first machine learning tutorial and have your experiment set up, you can
continue to improve the model and then deploy it as a predictive web service.
TIPTIP
Iterate to try to improve the model - For example, you can change the features you use in your
prediction. Or you can modify the properties of the Linear Regression algorithm or try a different
algorithm altogether. You can even add multiple machine learning algorithms to your experiment at one
time and compare two of them by using the Evaluate Model module. For an example of how to compare
multiple models in a single experiment, see Compare Regressors in the Azure AI Gallery.
To copy any iteration of your experiment, use the SAVE AS button at the bottom of the page. You can see all the
iterations of your experiment by clicking VIEW RUN HISTORY at the bottom of the page. For more details, see
Manage experiment iterations in Azure Machine Learning Studio.
Deploy the model as a predictive web service - When you're satisfied with your model, you can deploy it
as a web service to be used to predict automobile prices by using new data. For more details, see Deploy an
Azure Machine Learning web service.
Want to learn more? For a more extensive and detailed walkthrough of the process of creating, training, scoring,
and deploying a model, see Develop a predictive solution by using Azure Machine Learning.
Walkthrough: Develop a predictive analytics solution
for credit risk assessment in Azure Machine Learning
3/21/2018 • 2 min to read • Edit Online
NOTENOTE
The problem
The solution
In this walkthrough, we take an extended look at the process of developing a predictive analytics solution in
Machine Learning Studio. We develop a simple model in Machine Learning Studio, and then deploy it as an
Azure Machine Learning web service where the model can make predictions using new data.
This walkthrough assumes that you've used Machine Learning Studio at least once before, and that you have
some understanding of machine learning concepts. But it doesn't assume you're an expert in either.
If you've never used Azure Machine Learning Studio before, you might want to start with the tutorial, Create
your first data science experiment in Azure Machine Learning Studio. That tutorial takes you through Machine
Learning Studio for the first time. It shows you the basics of how to drag-and-drop modules onto your
experiment, connect them together, run the experiment, and look at the results. Another tool that may be helpful
for getting started is a diagram that gives an overview of the capabilities of Machine Learning Studio. You can
download and print it from here: Overview diagram of Azure Machine Learning Studio capabilities.
If you're new to the field of machine learning in general, there's a video series that might be helpful to you. It's
called Data Science for Beginners and it can give you a great introduction to machine learning using everyday
language and concepts.
You can try Azure Machine Learning for free. No credit card or Azure subscription is required. Get started now.
Suppose you need to predict an individual's credit risk based on the information they gave on a credit
application.
Credit risk assessment is a complex problem, but we can simplify it a bit for this walkthrough. We'll use it as an
example of how you can create a predictive analytics solution using Microsoft Azure Machine Learning. To do
this, we use Azure Machine Learning Studio and a Machine Learning web service.
In this detailed walkthrough, we start with publicly available credit risk data and develop and train a predictive
model based on that data. Then we deploy the model as a web service so it can be used by others for credit risk
assessment.
To create this credit risk assessment solution, we follow these steps:
1. Create a Machine Learning workspace
2. Upload existing data
3. Create an experiment
4. Train and evaluate the models
5. Deploy the web service
6. Access the web service
TIPTIP
You can find a working copy of the experiment that we develop in this walkthrough in the Azure AI Gallery. Go to
Walkthrough - Credit risk prediction and click Open in Studio to download a copy of the experiment into your
Machine Learning Studio workspace.
This walkthrough is based on a simplified version of the sample experiment, Binary Classification: Credit risk prediction, also
available in the Gallery.
Walkthrough Step 1: Create a Machine Learning
workspace
3/21/2018 • 1 min to read • Edit Online
TIPTIP
This is the first step of the walkthrough, Develop a predictive analytics solution in Azure Machine Learning.
1. Create a Machine Learning workspace
2. Upload existing data
3. Create a new experiment
4. Train and evaluate the models
5. Deploy the Web service
6. Access the Web service
To use Machine Learning Studio, you need to have a Microsoft Azure Machine Learning workspace. This
workspace contains the tools you need to create, manage, and publish experiments.
The administrator for your Azure subscription needs to create the workspace and then add you as an owner or
contributor. For details, see Create and share an Azure Machine Learning workspace.
After your workspace is created, open Machine Learning Studio (https://studio.azureml.net/Home). If you have
more than one workspace, you can select the workspace in the toolbar in the upper-right corner of the window.
If you were made an owner of the workspace, you can share the experiments you're working on by inviting others to the
workspace. You can do this in Machine Learning Studio on the SETTINGS page. You just need the Microsoft account or
organizational account for each user.
On the SETTINGS page, click USERS, then click INVITE MORE USERS at the bottom of the window.
Next: Upload existing data
Walkthrough Step 2: Upload existing data into an
Azure Machine Learning experiment
5/1/2018 • 3 min to read • Edit Online
Convert the dataset format
This is the second step of the walkthrough, Develop a predictive analytics solution in Azure Machine Learning
1. Create a Machine Learning workspace
2. Upload existing data
3. Create a new experiment
4. Train and evaluate the models
5. Deploy the Web service
6. Access the Web service
To develop a predictive model for credit risk, we need data that we can use to train and then test the model. For
this walkthrough, we'll use the "UCI Statlog (German Credit Data) Data Set" from the UC Irvine Machine
Learning repository. You can find it here:
http://archive.ics.uci.edu/ml/datasets/Statlog+(German+Credit+Data)
We'll use the file named german.data. Download this file to your local hard drive.
The german.data dataset contains rows of 20 variables for 1000 past applicants for credit. These 20 variables
represent the dataset's set of features (the feature vector), which provides identifying characteristics for each
credit applicant. An additional column in each row represents the applicant's calculated credit risk, with 700
applicants identified as a low credit risk and 300 as a high risk.
The UCI website provides a description of the attributes of the feature vector for this data. This includes financial
information, credit history, employment status, and personal information. For each applicant, a binary rating has
been given indicating whether they are a low or high credit risk.
We'll use this data to train a predictive analytics model. When we're done, our model should be able to accept a
feature vector for a new individual and predict whether he or she is a low or high credit risk.
Here's an interesting twist. The description of the dataset on the UCI website mentions what it costs if we
misclassify a person's credit risk. If the model predicts a high credit risk for someone who is actually a low credit
risk, the model has made a misclassification. But the reverse misclassification is five times more costly to the
financial institution: if the model predicts a low credit risk for someone who is actually a high credit risk.
So, we want to train our model so that the cost of this latter type of misclassification is five times higher than
misclassifying the other way. One simple way to do this when training the model in our experiment is by
duplicating (five times) those entries that represent someone with a high credit risk. Then, if the model
misclassifies someone as a low credit risk when they're actually a high risk, the model does that same
misclassification five times, once for each duplicate. This will increase the cost of this error in the training results.
The original dataset uses a blank-separated format. Machine Learning Studio works better with a comma-
separated value (CSV) file, so we'll convert the dataset by replacing spaces with commas.
There are many ways to convert this data. One way is by using the following Windows PowerShell command:
cat german.data | %{$_ -replace " ",","} | sc german.csv
sed 's/ /,/g' german.data > german.csv
Upload the dataset to Machine Learning Studio
Another way is by using the Unix sed command:
In either case, we have created a comma-separated version of the data in a file named german.csv that we can
use in our experiment.
Once the data has been converted to CSV format, we need to upload it into Machine Learning Studio.
1. Open the Machine Learning Studio home page (https://studio.azureml.net).
2. Click the menu in the upper-left corner of the window, click Azure Machine Learning, select
Studio, and sign in.
3. Click +NEW at the bottom of the window.
4. Select DATASET.
5. Select FROM LOCAL FILE.
6. In the Upload a new dataset dialog, click Browse and find the german.csv file you created.
7. Enter a name for the dataset. For this walkthrough, call it "UCI German Credit Card Data".
8. For data type, select Generic CSV File With no header (.nh.csv).
9. Add a description if you’d like.
10. Click the OK check mark.
This uploads the data into a dataset module that we can use in an experiment.
You can manage datasets that you've uploaded to Studio by clicking the DATASETS tab to the left of the Studio
window.
For more information about importing other types of data into an experiment, see Import your training data into
Azure Machine Learning Studio.
Next: Create a new experiment
Walkthrough Step 3: Create a new Azure Machine
Learning experiment
3/21/2018 • 6 min to read • Edit Online
This is the third step of the walkthrough, Develop a predictive analytics solution in Azure Machine Learning
1. Create a Machine Learning workspace
2. Upload existing data
3. Create a new experiment
4. Train and evaluate the models
5. Deploy the Web service
6. Access the Web service
The next step in this walkthrough is to create an experiment in Machine Learning Studio that uses the dataset we
uploaded.
1. In Studio, click +NEW at the bottom of the window.
2. Select EXPERIMENT, and then select "Blank Experiment".
3. Select the default experiment name at the top of the canvas and rename it to something meaningful.
Prepare the data
TIPTIP
4. In the module palette to the left of the experiment canvas, expand Saved Datasets.
It's a good practice to fill in Summary and Description for the experiment in the Properties pane. These
properties give you the chance to document the experiment so that anyone who looks at it later will understand
your goals and methodology.
5. Find the dataset you created under My Datasets and drag it onto the canvas. You can also find the dataset
by entering the name in the Search box above the palette.
You can view the first 100 rows of the data and some statistical information for the whole dataset: Click the
output port of the dataset (the small circle at the bottom) and select Visualize.
Because the data file didn't come with column headings, Studio has provided generic headings (Col1, Col2, etc.).
Good headings aren't essential to creating a model, but they make it easier to work with the data in the
experiment. Also, when we eventually publish this model in a web service, the headings help identify the columns
to the user of the service.
We can add column headings using the Edit Metadata module. You use the Edit Metadata module to change
metadata associated with a dataset. In this case, we use it to provide more friendly names for column headings.
To use Edit Metadata, you first specify which columns to modify (in this case, all of them.) Next, you specify the
action to be performed on those columns (in this case, changing column headings.)
TIPTIP
1. In the module palette, type "metadata" in the Search box. The Edit Metadata appears in the module list.
2. Click and drag the Edit Metadata module onto the canvas and drop it below the dataset we added earlier.
3. Connect the dataset to the Edit Metadata: click the output port of the dataset (the small circle at the bottom
of the dataset), drag to the input port of Edit Metadata (the small circle at the top of the module), then
release the mouse button. The dataset and module remain connected even if you move either around on
the canvas.
The experiment should now look something like this:
The red exclamation mark indicates that we haven't set the properties for this module yet. We'll do that
next.
You can add a comment to a module by double-clicking the module and entering text. This can help you see at a
glance what the module is doing in your experiment. In this case, double-click the Edit Metadata module and type
the comment "Add column headings". Click anywhere else on the canvas to close the text box. To display the
comment, click the down-arrow on the module.
4. Select Edit Metadata, and in the Properties pane to the right of the canvas, click Launch column
selector.
5. In the Select columns dialog, select all the rows in Available Columns and click > to move them to
Selected Columns. The dialog should look like this:
Status of checking account, Duration in months, Credit history, Purpose, Credit amount, Savings
account/bond, Present employment since, Installment rate in percentage of disposable income, Personal
status and sex, Other debtors, Present residence since, Property, Age in years, Other installment
plans, Housing, Number of existing credits, Job, Number of people providing maintenance for,
Telephone, Foreign worker, Credit risk
6. Click the OK check mark.
7. Back in the Properties pane, look for the New column names parameter. In this field, enter a list of
names for the 21 columns in the dataset, separated by commas and in column order. You can obtain the
columns names from the dataset documentation on the UCI website, or for convenience you can copy and
paste the following list:
The Properties pane looks like this:
TIPTIP
Create training and test datasets
If you want to verify the column headings, run the experiment (click RUN below the experiment canvas). When it finishes
running (a green check mark appears on Edit Metadata), click the output port of the Edit Metadata module, and select
Visualize. You can view the output of any module in the same way to view the progress of the data through the
experiment.
We need some data to train the model and some to test it. So in the next step of the experiment, we split the
dataset into two separate datasets: one for training our model and one for testing it.
To do this, we use the Split Data module.
TIPTIP
1. Find the Split Data module, drag it onto the canvas, and connect it to the Edit Metadata module.
2. By default, the split ratio is 0.5 and the Randomized split parameter is set. This means that a random half
of the data is output through one port of the Split Data module, and half through the other. You can adjust
these parameters, as well as the Random seed parameter, to change the split between training and testing
data. For this example, we leave them as-is.
The property Fraction of rows in the first output dataset determines how much of the data is output through
the left output port. For instance, if you set the ratio to 0.7, then 70% of the data is output through the left port
and 30% through the right port.
3. Double-click the Split Data module and enter the comment, "Training/testing data split 50%".
We can use the outputs of the Split Data module however we like, but let's choose to use the left output as
training data and the right output as testing data.
As mentioned in the previous step, the cost of misclassifying a high credit risk as low is five times higher than the
cost of misclassifying a low credit risk as high. To account for this, we generate a new dataset that reflects this cost
function. In the new dataset, each high risk example is replicated five times, while each low risk example is not
replicated.
We can do this replication using R code:
dataset1 <- maml.mapInputPort(1)
data.set<-dataset1[dataset1[,21]==1,]
pos<-dataset1[dataset1[,21]==2,]
for (i in 1:5) data.set<-rbind(data.set,pos)
maml.mapOutputPort("data.set")
1. Find and drag the Execute R Script module onto the experiment canvas.
2. Connect the left output port of the Split Data module to the first input port ("Dataset1") of the Execute R
Script module.
3. Double-click the Execute R Script module and enter the comment, "Set cost adjustment".
4. In the Properties pane, delete the default text in the R Script parameter and enter this script:
TIPTIP
We need to do this same replication operation for each output of the Split Data module so that the training and
testing data have the same cost adjustment. The easiest way to do this is by duplicating the Execute R Script
module we just made and connecting it to the other output port of the Split Data module.
1. Right-click the Execute R Script module and select Copy.
2. Right-click the experiment canvas and select Paste.
3. Drag the new module into position, and then connect the right output port of the Split Data module to the
first input port of this new Execute R Script module.
4. At the bottom of the canvas, click Run.
The copy of the Execute R Script module contains the same script as the original module. When you copy and paste a
module on the canvas, the copy retains all the properties of the original.
Our experiment now looks something like this:
For more information on using R scripts in your experiments, see Extend your experiment with R.
Next: Train and ev