Upcoming Webinar June 15: Translating Insight Into Action: Recommendations to Know, Win and Retain Customers

Close Message

# How Does AI Make Recommendations? An Easy-To-Understand Primer.

Artificial intelligence is used to make recommendations to customers. Learn how recommendations are created by computer algorithms in a simple way you can understand.

You can binge-watch both  season one and season two right now.

Subscribe to the Lucid Thoughts channel and be sure to leave your questions and comments on each video.

Transcript:

Welcome back, my friend.

Let’s talk recommendations.

You’ve probably made a ton, which restaurants people should try, which music they should check out, where they should go on their first date. Funny that one didn’t catch on. You shouldn’t be surprised to learn that we can teach computers to make recommendations too.

The first thing you need to know is your recommendations are only as good as your data, and you need lots and lots of data.

For an easy explanation, let’s take this back to our playground days, shall we?

Emily loves spending her time both on the swings and the see-saw. Bianca also loves the swings and the see-saw. That means we can group Emily and Bianca together, and when the new girl Joanna joins the class, she tells everyone she loves the swings. This puts Joanna near Emily and Bianca’s group, so it’s safe to assume we can recommend the see-saw to Joanna.

Follow? Good.

Now, let’s get a little bit more technical, let’s really delve into this algorithm to pinpoint the precise nuisances of it.

Just kidding. That was an alternating least squares formula, and while it is important, it won’t be on the test.

Basically, it’s a formula that helps you get nifty charts like these.

Let’s say this chart or matrix, represents Kevin and his movie-viewing habits. He rates horror movies and documentaries pretty high. This one over here is Jake, Jake has watched similar movies which means Kevin and Jake are clustered together. But notice how Jake is missing a few titles that Kevin has seen.

Now, by multiplying these matrices together. Don’t worry, someone else is doing that right now. Then we get these new data points to fill in the empty squares for Jake. If we plot these data points, we get a line chart that helps us make predictions. That’s the heart of linear regression.

The closer a data point is to Jake’s line, the better the recommendation is for him. So obviously, the machine will recommend Texas Chainsaw Massacre over Merry Kissmas.

Now, if the computer doesn’t have enough data to make a prediction, you get what’s called a sparse matrix. Now if you try to force this data into this scenario, you’ll probably get some bad recommendations.

Everyone wants to tell you which new show you should binge watch next, including computers, but these machines have the data to back it up, so you know their recommendations are legit.

That’s why tonight I can’t wait to dive into Bridget Jones’s Diary.

Whoa, whoa! Come on guys, it’s not funny.