Hear from Katie Schuler, Software Developer at Network Center, Inc. about the advent of Machine Learning and how implementing it into your everyday business practices is a lot more cost-effective and realistic than you may have initially thought.
Earlier this month I attended a live webinar presented by Microsoft called “Azure Data Analytics for Developers”. One of the main topics was Machine Learning, which I was under the impression is a technology that only large companies could afford to develop and use. I was surprised to learn how easy and relatively inexpensive these technologies can be to implement. Let’s take a look at what Machine Learning could mean for you.
What is Machine Learning?
The technical definition includes words and phrases like artificial intelligence, pattern recognition, algorithms, etc. But the basic concept of machine learning is computers that are setup to learn and make predictions without explicit human programming.
You’ve likely already seen this technology in action. Machine learning is how online shopping sites recommend other products after a user adds an item to their cart. Spam filters can use machine learning to determine which emails should be prevented from reaching your inbox. Credit card companies use machine learning to detect fraudulent charges. Social media sites use it to help you connect with other people, companies, or interests that are most relevant to you. Machine learning also has applications with regard to object recognition for cameras, medical diagnosis, stock market analysis, language and handwriting recognition, and countless other examples.
How can it help you?
First and foremost, you need to figure out where in your business processes machine learning can benefit you. Machine learning is most advantageous with a specifically defined question or issue in mind, albeit an issue where the path to the outcome is not easily reached by human means.
A common example is searching for credit card fraud. The question/issue is specific and clear: Is a charge of $55 at a convenience store in Washington D.C. fraudulent or not. But the path to the outcome is not necessarily easily reached. It involves many different variables. Has the cardholder shopped there before? Does the cardholder live in Washington D.C.? If not, has the cardholder made a purchase from an airline in the last few months or had charges in an airport recently? These and many other factors that contribute to this scenario would take a human several minutes, at best, to decide if the charge should be flagged as suspicious or not.
How does it work?
There are many different software options out there for setting up your company to use machine learning. All of them will have similar steps to get them working accurately for your company. The webinar I attended was specifically for Microsoft Azure, which has a slick and easy-to-use flowchart style interface. Users simply drag and drop different components into a workspace, connect them, set a few properties, and let it run. Here are the basic steps:
- Connect the software with historical data from your database that includes outcomes of what you want to later predict for new data.
- Connect an algorithm type, which depends on the data and the question at hand.
- It may take some trial and error to find the best fitting algorithm for the data.
- Train the model to predict outcomes.
- This is where the software will take the historical data and outcomes and learn from it to create a model.
- Score and evaluate the model.
- This requires another portion of historical data, including outcomes, which the software will use to determine how accurate it is at predicting the correct outcome.
- If the score is low, you may need to try a different algorithm type.
- Once you’re satisfied with the model, it can now be applied to new data to predict outcomes.
A few simple steps and you’ve got information that can give your company an edge on the competition.
For more information about Microsoft Azure Machine Learning, checkout this free video course at the Microsoft Virtual Academy:
Contact Network Center, Inc. for further details on how you can use machine learning to enhance your data usage.