Skip to content

1. Introduction to Make a Recommendation
attachment
Students can submit work

Playback Speed:
Transcript

In this lesson, you will create a recommendation system for items in the category you selected.

To do this, you will develop a model.

A “model” is a system or instruction for completing a task or solving a problem.

More and more, models are being created through machine learning.

Computers use these models to: check your spelling, make calculations, search the internet, and suggest the next word or phrase to write in an email.

The model you develop will help you predict how much someone will like an item based on how much they liked other, similar items.

By being able to predict a rating, you can recommend more items that person will like.

Imagine going shopping with a longtime friend.

People who have been friends for many years often enjoy the same activities, hobbies, and products.

So, if your friend sees an item that they enjoy, they may recommend it to you.

Often, you will like what they recommend, too.

A computer can’t build friendships in this way, but it can identify people who like the same things as you.

Computers have learned that people with similar interests often enjoy the same things.

So, they use data to develop models to help make these predictions.

The more data a computer has about someone’s likes and dislikes, the more accurate the recommendation system becomes.

In this lesson, you will use the data about likes and dislikes that you collected from your survey.

Then, you will identify patterns in the data.

Sometimes data does not have a pattern, and that’s okay, too.

To build your model, you will calculate the difference between the rating a user gave the first item in the form...

And every other item in your form.

Then, you will use the average of those differences for all of the people who provided response data.

The average differences enable you to determine what someone would rate the first item in your form if you had not collected that data from them.

For example, think of a survey about different flowers.

Someone rates a tiger lily a three, on a one-to-five scale.

If you identify a pattern in the data showing that the average user rates sunflowers one point higher than tiger lilies, then you can add the average differences of the known ratings to predict that the person will rate the sunflower a four.

Computers identify patterns in data in the same way.

However, the models they create are not always accurate.

For instance, just because most users, on average, prefer sunflowers, doesn’t mean it’s always true, or that you can rely on that result.

For instance, if someone happens to be allergic to sunflowers, they will probably give them a low rating, even though the data predicts something different.

In the same way, the model you develop may not be successful.

That’s okay.

To get started, open the spreadsheet you created in the previous lesson.

If you do not have a spreadsheet with collected data, open the starter project and make a copy.

Then, move on to the next video to start analyzing responses.

Now, it’s your turn: Open your form responses spreadsheet, Or open the starter project and make a copy.

Next
Instructions
  1. Open your form responses spreadsheet.
  2. Or, open the starter project and make a copy.