>>Hi, my name is Eduardo Mellow, and I’m a program manager on
the Azure Machine Learning Team. Today I want to show you some of the new capabilities we’re adding to the Azure Machine Learning
Service that makes it very easy for you to
create experiments, train models, test them, and deploy the models to production, without having to write
a single line of code. Let’s take a look at that in action. In this example, I want to create a model that is able to
predict the price of a car based on some properties
such as mileage per gallon, horsepower, and whether it
has four-wheel drive or not. In the experiment, I start with an existing dataset
with historical data, and I already have
a few steps set up here. One that keeps only the columns that are relevant to train the model, I have another step that we remove
all the rows with missing data, and one final step
that splits the data into a training dataset
in a test data set. Now, let’s add
a few more steps so that we can train a machine learning model. Since this is a prediction problem, I will be using a regression
approach to these, specifically, I’m going to use linear regression, there are a few more options
that I can choose from as well. Another step that I need to add
to the experiment is to train the model using the training dataset. I also need to specify
for the model training, which column the model
is trying to predict. In this particular case, the model is trying to
predict the price of the car, so that’s the column
that I need to select. I need to add another step
true scored the model in that so that we can
see the results of the training model
using the test data. One final step, to evaluate
the overall result of the model. So these are all the steps it takes to train a machine learning
model into end. As we’ve seen, it was
a very simple experience, we’ve just drag and drop
the relevant components and connecting them together
to create these experiments. I click “Run” I select one of the existing compute targets that is available already in my environment, and within a few minutes, I should have my first trained
machine learning model. Now that the model has been trained, let’s look at the results. I can click on “Score
Model” and choose to see the results from
the test dataset. These will show me all the
samples from the test dataset, including the expected price
for those cars, and also the model prediction, meaning the price that the model predicted for
that particular car, and I can compare them side-by-side. A more interesting way to look
at the quality of the model is to click on the “Evaluate Model”
and click on “Visualize”. This will show
key performance metrics that indicates the precision and
the error rates of that model. In this particular case, I can do much better
by training the model with more data or by using
different regression models. However, this gives you
a complete end-to-end scenario for how you can train models in
Azure Machine Learning Service. I will also point you to a set
of samples that we provide that give you more complex
experiments with different models, including multiple models
in the same experiment, and that will give you
a sense of all that you can do with Azure Machine
Learning Service. Deploying that model to production
is also just a few clicks. Once deployment is complete, I have a web endpoint that I can use to get predictions
based on that model. There are few things that
I want to show here that I think you’re going to
find also very interesting. First, if I click on “Consume”, I can get C-sharp, R, and Python code samples that I can simply copy and paste
into my application, and that will call the model
as part of the application. It’s that simple tree incorporates that machine learning model
into your application. Moreover, I can go to “Test” and test the model deployed as
a web service, from right here. All I need to do is specify
the input values and click “Test”, that will send a request to that web endpoints that is
hosting the model, and I can get the results back
in terms of the input that I provided as well as the prediction from the model
that just got deployed. This is the power of Azure
machine learning that you can have in train models without
writing any line of code. I hope you’re going to find
these useful and compelling, and you’ll give it a try yourself. [MUSIC].

Build “zero code” machine learning models with Azure Machine Learning service
Tagged on:                         

2 thoughts on “Build “zero code” machine learning models with Azure Machine Learning service

  • May 3, 2019 at 1:49 am

    Would like to see "Julia-Lang" samples too.

  • May 4, 2019 at 5:46 pm

    Alfa Rome-R-o XD


Leave a Reply

Your email address will not be published. Required fields are marked *