Using Netflix To Make Sense Of Machine Learning

Machine learning is one of the most common buzzwords thrown around the world of digital marketing today. For all of the powerful and amazing capabilities this data science function possess, it grinds my gears just having to use it in this post. I want to attempt to demystify this played out phrase, by showing that anyone can create a “machine learning algorithm” in 15 minutes.

The most widespread usages of machine learning that virtually everyone touches on a day to day basis are recommendation engines, which use pre-existing behavior data to predict future actions. Think of a time when Amazon was able to predict that you would need a helmet to go with your shiny new goggles, or when Netflix knew that Pulp Fiction was perfect for your binging marathon of Tarantino films. These are all examples of the most core principles that machine learning provides.

In its simplest form machine learning is creating and training models that are learned from raw data sets.

Typically, our goal will be to use existing data to develop models that can predict outcomes for a behavior/output, such as what movies you would enjoy watching on Netflix.

This very basic exercise from a Siraj Naval’s coding challenge video helps us to begin to work through the most basic forms of machine learning. Using numpy to handle the raw data and LightFM (a basic recommendation engine), we can walk through the steps of creating our own personal Netflix-esque algorithm in a flash. A sample output from this challenge would look something like this:

In under 50 lines of code, we can import some data (Netflix movie catalog), train the data based on pre-existing variables(likes/ratings etc.), and print an output that allows you to recommend a movie an individual user in a database would like to watch.

The actual algorithm that powers Netflix’s recommendation engine is infinitely more nuanced, but the point is that the basic steps in training a computer to predict future preferences intelligently are a perfect example of how machine learning functions in our day to day lives.