By: Jason Laverty
28 July, 2015
Machine Learning: A complex beast, easy to tame.
Last week we took a quick glimpse at how machine learning is being used in web development, specifically its application to ecommerce.
We've had some feedback from the post that people are interested in giving machine learning a crack, but have no idea how to start, so we're about to mix it up and go full tutorial mode - I'm talking over 9000.
Most web developers have a few tricks up their sleeves when it comes to programming, a bit of PHP here, Ruby on Rails there, some even prefer to get nice and cosy with Microsoft and dabble in .NET. Most off the shelf frameworks we reviewed are created in Java or Python - so if you're familiar with these you already have a foot in the door. For those who are not it's the mega giants of tech to the rescue in the form of Google, Amazon, Microsoft and IBM.
You're probably thinking, what? Amazon? I thought they only sell books?! Amazon has come a long way since providing bedtime reading on your Kindle. Amazon along with the other aforementioned companies have done all the hard yards so that you don't have to, exposing easy-to-use APIs and creating user-friendly interfaces, allowing you to get started with Machine learning now - with the languages you already know and love.
How does it work with your existing language? We'll technically it doesn't, but mining your own information, and having that ready to post to the machine means you're already half way there. The rest requires usually posting that information to a server via a CVS file, old-school, or via a REST API using data markup such as JSON or XML, the later is particularly important if you're looking to contstantly feed the machine tasty morsels in the form of information cookies.
Your ecommerce website is an ever evolving beast, constantly churning new information as each new customer arrives at your doorstep - you'll want to feed all that new information back into the machine in the form of training. Theres little point training a machine on volumes and volumes of information just to draw a line in the sand and say done. Your predictions will become increasing accurate, and if your information catchment has been designed to put forward as many different types of attributes as possible, you may even be able to create new types of predictions from the same trained set.
Now that you know that machine learning is within your everyday grasp, it's time to tackle those four tech giants head on and find out who is offering what, how easy they are to use and what the prices are.
Next up: Part two of the series - Machine Learning: The Four Way Brawl.