Code Library

One of the most talked about emerging technologies in the last few years is machine learning. Machine learning isn’t new by any stretch; however, the ability for a wide range of people to easily access the technology is relatively new—largely enabled by the ubiquitous nature of public cloud computing.

By using machine learning, we are able to leverage computers to analyze existing data to predict future behavior or outcomes. Maybe you want to predict when machines will break down instead of sending technicians to check on machines that may be working fine. Perhaps you want to analyze historical shopping data to predict what customers are likely to purchase in the future. These are just two of the many potential scenarios that are possible with machine learning.

It’s becoming increasingly common to see machine learning mentioned along with Internet of Things (IoT). IoT solutions typically generate a lot of data from various sensors, such as temperature, vibrations, speed, and so on. The data itself is often largely useless—the true value comes in being able to analyze the data and determine what to do with the data to improve the overall solution. This is where machine learning comes into the picture. Combining the data from IoT solutions with machine learning can lead to interesting and useful insights about the data.

At first this might seem daunting, but it isn’t—especially when using a service such as Azure Machine Learning. With Azure Machine Learning, there is no hardware to purchase or virtual machines to manage. In fact, you don’t even need an Azure subscription to get started with the Free tier of Azure Machine Learning! You can learn more about the pricing options for Azure Machine Learning at https://azure.microsoft.com/pricing/details/machine-learning/.

The basic workflow in Azure Machine Learning is relatively simple:
1. Build a model from existing data. The data can come from numerous data stores in Azure, such as Azure Storage tables or blobs, Azure HDInsight (Hadoop), Azure SQL Database, or Azure Data Lake.

2. Publish the model as a web service.

3. Optionally, consume that web service from any number of tools such as mobile applications, websites, or business intelligence tools.

With the data in place, you can create your predictive model in Azure Machine Learning Studio, a browser-based tool with drag-and-drop capabilities that make it easy to get started. If you’re familiar with machine learning and understand how to use R (a programming language commonly used for data analysis) or Python, you can get started right away with Azure Machine Learning. If you’re not familiar with machine learning, you can get started by using solution templates in the Cortana Intelligence Gallery or by leveraging existing solutions in the Azure Marketplace (Data Market).
Once the model is created and properly trained (a process for validating that the model works as expected), you can publish the model as a web service. This will allow you—or others, based on your usage needs—to send data to your service and receive the predictions!

Azure Machine Learning is the perfect complement to the voluminous amount of data generated by many of today’s IoT solutions. It’s never been easier to gather, store, analyze, and make decisions based on data.

Source of Information : Microsoft Azure Essentials Fundamentals of Azure Second Edition