Considering machine learning? Think about the problem you're solving first

Machine learning gets a lot of buzz, but United Shore's Brian Korzynski pointed out that it's only a good solution in certain situations, and it's important for companies to consider their needs.

Is machine learning right for your business?

Speaking with TechRepublic's Alison DeNisco Rayome at the recent Code PaLOUsa conference in Louisville, Brian Korzynski, senior application developer at United Shore, explained why it's essential that businesses understand the pros and cons of machine learning.

Read Korzynski's comments below, or watch the video:

Korzynski: There's been a lot of buzz in the news about what machining learning is and what it's not, and there's a lot of misconceptions that people get about what it really is. What it really is is a very specific set of algorithms that solve very specific problems. Machine learning is only targeted towards certain types of problems such as classifications, or regressions, recommendations, those kinds of things.

The subset is a lot smaller than what a lot of people think, but a lot of it gets misconstrued, based upon a lot of the things you see in the news, such as like self-driving cars, and how machines are generating scripts for movies, and things like that.

SEE: Managing AI and ML in the enterprise (ZDNet special report) | Download the report as a PDF (TechRepublic)

The advice that I'd have for people that want to get into machine learning is, first of all, you need to look at your business and what types of problems you're trying to solve to see if machine learning really fits that problem. It doesn't fit every problem, and the reason being is, let's say for example you have a use case that's solved by really well- known and well-documented processes. Introducing machine learning could introduce variability, where you don't necessarily want it. The biggest thing I would say is, A, figure out what types of problems you're trying to solve, more so than trying to match buzzwords, and B, figure out like what types of algorithms and stuff solve them, and then start targeting that as your entry point.

There's been a lot of good resources out there, especially from companies like Microsoft with Azure. There's a lot of good cheat sheets, if you will, that'll walk you through the different types of algorithms and things you can do with it. I'd recommend starting at like one of those as a source to figure out what types of problems you can actually solve.

Also see

machine learning
Image: iStock/NicoElNino