In this Azure tutorial, we will discuss Microsoft Azure Machine Learning Tutorial.
Table of Contents
- Azure Machine Learning Tutorial
- What is Machine Learning?
- Benefits of Azure Machine Learning | Why Azure Machine Learning?
- Difference between Azure Machine Learning Studio and Azure Machine Learning Service
- Working with Azure Machine Learning Studio
- Hands-on with Azure Machine Learning Studio
- Azure Machine Learning Designer
Azure Machine Learning Tutorial
Before moving to the actual topic, i.e., Microsoft Azure Machine Learning Tutorial, We should be aware of What is Machine Learning?
What is Machine Learning?
Machine Learning is the part of Artificial intelligence (AI) that allows the systems to learn without being explicitly Programmed. This is an application of Artificial intelligence.
Now the next question that comes to your mind is, What is Artificial intelligence (AI) then?
Artificial intelligence (AI) is the process of simulating a human’s intelligence, thinking, and behavior in the machine with the help of Programs.
Now, let’s move toward the Azure ML Tutorial.
Azure Machine Learning is a different set of services and tools that helps the developers to develop and deploy the different Machine learning models easily. Primarily, it also saves a lot of cost and time needed for the implementation.
The most interesting fact is that you can now create the machine learning model with a few clicks. Can you believe this? Yes, this is the truth; with the help of IDE, i.e., Azure Machine Learning Studio, we can easily develop and deploy the machine learning model.
Benefits of Azure Machine Learning | Why Azure Machine Learning?
- Azure ML provides many benefits, like saving lots of cost and time so the medium size and small size businesses can dare to start developing machine learning models. So, it allows different middle-scale and small companies to go with the Machine learning models.
- The other benefit of Azure ML is you can directly connect with the Azure SQL and other services, making the task for the developer easy.
- Azure Machine Learning supports the operation with massive data or Big data. There are no restrictions on the quantity of data.
- The pricing model is flexible for pricing with the “pay as you go” model, which helps the organization save a lot of costs.
- Azure ML is very user-friendly and comes with a simple set of tools that are easy to use.
- Azure Machine Learning Studio helps us with drag-and-drop features where no coding knowledge is required.
- Azure Machine Learning Studio has some in-built algorithms and data transformation tools that greatly help.
- The set of Azure ML tools is built with the latest technologies and features that help to provide more accurate predictions.
- It helps with real-time predictions with the fastest algorithm.
- We can Publish our data model as a web service, which helps.
- The clutter feature in Office 365 is also being used along with Azure Machine Learning, which helps limit the data’s inaccuracy.
Azure Machine Learning comes with the Azure Machine Learning Studio and Azure Machine Learning Service.
Azure Machine Learning Studio is the IDE (Integrated Development Environment) that helps us with the environment and options to develop the Machine learning models with the drag-and-drop facilities. The best part is no coding knowledge is required here.
Azure Machine Learning Service is a fully managed cloud service that helps us provide the facility to train, deploy, and manage machine learning models. Here, coding is involved. It supports open-source technologies.
You can use many open-source Python packages like TensorFlow, Pytorch, etc, along with the Azure Machine Learning Service.
Azure Machine Learning service also comes with the support with Visual Studio Code extension that helps you manage your resources and deployments in Visual Studio Code.
Difference between Azure Machine Learning Studio and Azure Machine Learning Service
Below are a few key differences between the Azure Machine Learning Studio and the Azure Machine Learning Service.
|Azure Machine Learning Studio
|Azure Machine Learning Service
|The IDE (Integrated Development Environment) for the development of the Machine learning model.
|Azure Machine Learning Service is a fully managed cloud service to manage the Machine Learning model.
|No coding needed.
|It is a coding environment.
|Provides lots of inbuilt Machine learning algorithms that you can use to build your Machine learning models.
|In case of any customization that is not supported by the Out Of Box approach with Azure Machine Learning Studio, you can go with the Azure Machine Learning Service.
|Provides the feature with a drag and drop facility to build the Azure Machine learning model.
|Provides the feature with a drag and drop facility to build the Azure Machine learning model.
These are a few key differences between Azure Machine Learning Studio and Azure Machine Learning Service.
Working with Azure Machine Learning Studio
Suppose we will discuss little more on the Azure Machine Learning Studio a little more. In that case, Microsoft Azure Machine Learning Studio is a tool that provides you the drag and drop feature and is used to build, test, and deploy the predictive analytics solution on your data.
As already discussed, We can publish Azure Machine Learning as a web service that can be consumed by different external applications, Excel, etc.
Developing an Azure Machine learning model is an iterative process where you use data from different sources, transform and analyze that data through various data manipulation mechanisms, and generate a set of Outputs.
Azure Machine Learning Studio provides a visual workspace that helps you easily build, test, and iterate the Predictive data analysis model. You can drag and drop the datasets and modules and then, after connecting them, form an experiment. You can use that experiment in your Azure Machine learning studio.
A valid experiment has the below characteristics
- The experiment should contain a minimum of one dataset and one module
- Datasets may be connected to only different modules.
- The modules can be connected to different datasets or different other modules.
- You have to set all the required parameters for the modules you use for the experiment.
A dataset is data that you will have to upload to Azure Machine Learning Studio that you will use for the data analytics model. You can upload the dataset as per your requirements. Along with this, you will also get a set of sample datasets from different areas. Those data sets you can utilize as per your requirement.
Some of the sample datasets available are
- MPG Dataset for various automobiles
- Weather Dataset
- Flight Delays Data
- Time Series Dataset
- Book Reviews from Amazon
- Movie Tweets
- Airport Codes Dataset
Modules are a different set of algorithms you perform on your data. Azure Machine Learning Studio has a different set of modules, from the data ingress functions to training, scoring, validation processes, etc.
Below are a few examples of the inbuilt modules.
- Linear Regression
- Score Model
- two-class Logistic Regression
There is no programming knowledge required while using Azure Machine Learning Studio. What you need is, you need datasets and modules to construct your predictive data analytics model.
Now, the next question that comes to mind is from where you will get the Azure Machine Learning studio.
Hands-on with Azure Machine Learning Studio
Before accessing the Azure Machine Learning Studio, As a prerequisite, what we need is an Azure account or Subscription. Make sure you have an Azure Account or Azure Subscription. If you don’t have an Azure account, you can Create An Azure Free Account now.
Now, assuming you have a valid Azure account, to access the Azure Machine Learning Studio, you can search for the Azure Machine Learning Studio in google and access the first link once the link opens, click on the Sign In button and enter your Azure credentials when it prompts you to enter your credentials.
Once you Sign In with your Azure credentials, you can see the Microsoft Azure Machine Learning Studio below.
Now, the next step is to select the Experiment from the left navigation and then click on the + New button to create a new experiment
Now, click on the Blank experiment as highlighted below.
You can see “Experiment created on the date.” Now, you can rename it to the Experiment name you want to mention. Here, I mentioned the experiment named Prediction for Diabetes.
Now, the next step is to select the DataSet. To do that, click on the Saved Datasets from the left side menu and Expand the Samples. You can able to see many sample datasets. From those datasets, select the dataset named Pima Indian Diabetes Binary Classification Dataset.
Now drag the Pima Indian Diabetes Binary Classification Dataset to the center of the screen as mentioned with the Arrow mark with the hints. Drag items here.
Now, if you want to see what exactly is inside the data set we have selected for our experiment, click on the 1 with circle option as highlighted below and then click on the Visualize option. You can also download that data by clicking on the Download link.
You can see now, it is showing the detailed data on that.
The next step is to select the set of columns you want to choose for your Data Model.
Now, on the left side, you can see a search bar. Search for Select Columns in Dataset, and drag that search result towards the center of the screen down to the Pima Indian Diabetes Binary Classification Dataset box Now connect both boxes by dragging from the dot from the first box. Once you connect both, it will highlight with an arrow, as shown below.
Now click on the Launch Column Selector button from the right side.
Now you can see it is shown with lists of 9 Columns. All are relevant to our topic, so you can select all the columns and click the > button. You can see that all the columns are now moved to the Selected Columns section. Now click the tick mark button on the bottom of the screen, as highlighted below.
Now the next step is, we need to split our data into the training and test datasets. To do that, Search for the Split Data from the left side navigation search bar.
Drag the search result Split Data on the Workspace under the Select Columns in the DataSet box. Then, connect the Select Columns in the DataSet box and the “Split Data” box by dragging the dot symbol highlighted below. Once both are connected, you can see an arrow as highlighted below. From the right side, you can change the percentage if you want.
Now, we need the algorithm to train our data model next. So many algorithms are available, but here, we will choose the two-class logistic regression algorithm. Now, you might be thinking about why this algorithm. OK, the answer here is this algorithm is used to predict the probability of an outcome.
We are splitting into two Outcomes here, using the two-class logistic regression algorithm.
It would be best if you searched for the two-class logistic regression from the left side search box, and then you will find the two-class logistic regression algorithm as the search result. Drag that algorithm to the workspace area as highlighted below.
Now that we have selected our algorithm, we must train our Data Model. For that what we will have to do is, we need to search for the Train Model and drag it to the workspace as highlighted below.
Now connect it from two-class Logistic Regression to the Train Model and Split Data to the Train Model as highlighted below. Now click on the Launch column selector to select the columns and select the first column. You can select any of the columns based on your requirements, I have selected the first column.
Now, we need to score our trained Model. To do that, Search for the Score Model and drag the search result to the Workspace. Connect it to the Train Model and then connect the Training Dataset from the Split Data to the Train Model.
The next step is to search for the Evaluate Model from the left navigation search area and then drag the result to the Workspace. Connect the Score Model with it. See the image below for your reference.
The next step is to click on the save button to save the experiment and then click on the Run button to run the experiment.
After you click on the run button, it will verify each stage and show a green tick mark on each stage. See the screenshot below for your reference. Once it finishes running, it will show Finished running.
Now click on the dot on the Evaluate Model and then click on the Visualize option to see the report.
You can see the report below. You can adjust the threshold to minimize the false negative value. Ideally, False Positive and False Negative should be minimum.
This is just one example of working with Azure Machine Learning Studio. You can do so many things using this.
Azure Machine Learning Designer
Azure Machine Learning designer helps you to connect the datasets and the modules in an interactive way that helps to create the Machine learning models.
The Azure Machine Learning Designer uses the Azure Machine Learning workspace to Organize the resources.
With the help of the Azure Machine Learning Designer, you can get a canvas to build, test, and deploy machine learning data models. You can perform the following activities
- Drag and drop the datasets and modules into the workspace or canvas area.
- You can connect with different modules and datasets with the modules, etc.
- You can submit a pipeline run using different Azure Machine learning workspace resources.
- It helps us to make predictions on new data in real-time by deploying a real-time inference pipeline to a real-time endpoint.
You may also like following the articles below
In this article, we discussed a complete step by step Azure ML tutorial.
I am Bijay, a Microsoft MVP (10 times) having more than 17 years of experience in the software industry. During my IT career, I got a chance to share my expertise in SharePoint and Microsoft Azure, like Azure VM, Azure Active Directory, Azure PowerShell, etc. I hope you will learn from these Azure tutorials. Read more