Chapter 10
I
K-nearest neighbors
It just involves a set of five numbers instead of a set of two numbers.
The distance formula is flexible: you could have a set of a
million
numbers and still use the same old distance formula to find the
distance. Maybe you’re wondering, “What does
distance
mean when
you have five numbers?” The distance tells you how similar those sets of
numbers are.
Here’s the distance between Priyanka and Justin.
Priyanka and Justin are pretty similar. What’s the difference between
Priyanka and Morpheus? Calculate the distance before moving on.
Did you get it right? Priyanka and Morpheus are 24 apart. The distance
tells you that Priyanka’s tastes are more like Justin’s than Morpheus’s.
Great! Now recommending movies to Priyanka is easy: if Justin likes a
movie, recommend it to Priyanka, and vice versa. You just built a movie
recommendations system!
If you’re a Netflix user, Netflix will keep telling you, “Please rate more
movies. The more movies you rate, the better your recommendations
will be.” Now you know why. The more movies you rate, the more
accurately Netflix can see what other users you’re similar to.
195
Building a recommendations system
EXERCISES
10.1
In the Netflix example, you calculated the distance between two
different users using the distance formula. But not all users rate
movies the same way. Suppose you have two users, Yogi and Pinky,
who have the same taste in movies. But Yogi rates any movie he
likes as a 5, whereas Pinky is choosier and reserves the 5s for
only the best. They’re well matched, but according to the distance
algorithm, they aren’t neighbors. How would you take their
different rating strategies into account?
10.2
Suppose Netflix nominates a group of “influencers.” For example,
Quentin Tarantino and Wes Anderson are influencers on Netflix,
so their ratings count for more than a normal user’s. How would
you change the recommendations system so it’s biased toward the
ratings of influencers?
Regression
Suppose you want to do more than just recommend a movie: you want
to guess how Priyanka will rate this movie. Take the five people closest
to her.
By the way, I keep talking about the closest five people. There’s nothing
special about the number 5: you could do the closest 2, or 10, or 10,000.
That’s why the algorithm is called k-nearest neighbors and not five-
nearest neighbors!
Suppose you’re trying to guess a rating for
Pitch Perfect
. Well, how did
Justin, JC, Joey, Lance, and Chris rate it?
|