Share →
Buffer

Azure Machine Learning

It’s everywhere. Your email system likely uses it for spam identification. Maps uses it to provide you the most optimal, up-to-date route to a given destination. Cortana and Siri use it to better understand how you talk. Facebook uses it to prioritize delivery of what’s in your news feed. The list goes on and on. Machine learning (ML) is being used to optimize, prioritize, and predict behavior, both human and systems. But what is it, and how can it help your organization?

Wikipedia states, “Machine learning is a scientific discipline that explores the construction and study of algorithms that can learn from data. Such algorithms operate by building a model from example inputs and using that to make predictions or decisions, rather than following strictly static program instructions.” To put it more precisely, with machine learning, computers are given the ability to learn without being explicitly programmed. Machine learning is one facet of artificial intelligence.

At risk of over-simplifying how it works, machine learning takes a series of data and uses a set of rules and algorithms to parse through the data, creating models that are then used to provide output. What makes machine learning so powerful is that while it is parsing through each set of data, it is using the new data to adjust what it has learned from the previous set of data, thus (potentially) resulting in progressively more accurate predictions.

For example, let’s say you’re building a two-legged robot that is supposed to walk. You could program it to walk, helping it to understand various topographies and how it should react, but the problem lies in the unexpected. You will likely not be able to program for every scenario and situation. What happens if you build another robot? Maybe the second robot’s center of gravity is slightly off, or the servo’s you’ve built are a different make, resulting in a different rate of movement. There are many situations you need to consider, yet it is an unmanageable situation because it’s impossible to create programs and sub routines to address every single scenario. This is where machine learning becomes powerful. Machine learning gives the walking robot the ability to fail and then try something different until it overcomes the challenge. It then uses these lessons and will apply them as the situation demands. With machine learning, you’d start the robot on a flat, stable surface, allowing it to learn how to walk. Then you introduce different environmental challenges. As it encounters them, it will use its base routines to try and adapt to the different surfaces, learning through success and failure. Eventually it creates models that will allow it to successfully navigate these new topographies.

Similarly, consider a situation where you want to identify which restaurants you’d most likely enjoy. You, of course, know instinctively what they could be, but the computer does not. And of course, you don’t want to sample every single restaurant in the city just to figure out which ones you like. Machine learning can take a set of inputs such as type of cuisine, price, location, atmosphere, average of clientele, rating, size, and even date opened, and predict which restaurants you are more likely to enjoy. You will need to provide some initial inputs so it can learn which ones you have experienced and liked or disliked, and then, over time and more feedback, it can, with increased accuracy, provide recommendations. And since this idea might have you salivating, there’s an Azure ML demonstration that shows how to use the Matchbox recommender module to train a restaurant recommendation engine. Bon Appetite!

When it comes to using machine learning for your own business, the sky is the limit. Machine learning can act as an oracle when it comes to making business decisions, provided you have data to help guide the initial learning. In Azure Machine Learning there are many sample experiments to get you started. Check out the E-mail classifier training, the predictive maintenance model, and the Twitter sentiment analysis experiments as examples of some of the ways you can use Azure Machine Learning. These experiments are just a few of many you can use. After you’ve spent a little time with these, you’ll be ready to start playing with your own experiments and developing models using your own data!

Azure Machine Learning has the ability to open up the mysteries locked in your data. You’ve been gathering this data forever and I’m sure you have questions that you know your data can answer, but you just don’t know how to get it to tell you. Machine learning may be the answer.

Get started by visiting the Azure Machine Learning page. I also recommend checking out this insightful paper that offers a Few Useful Things to Know about Machine Learning.

Print Friendly