Non-personalized recommendations

Building Recommendation Engines in Python

Rob O'Callaghan

Director of Data

Non-personalized ratings

Example of three books often bought together.

Building Recommendation Engines in Python

Finding the most popular items

book_df DataFrame:

User Book
User_233 The Great Gatsby
User_651 The Catcher in the Rye
User_131 The Lord of the Rings
User_965 Little Women
User_651 Fifty Shades of Grey
... ...
Building Recommendation Engines in Python

Finding the most popular items

book_df['book'].value_counts()

40 Shades of Grey                      524
Harry Potter and the Sorcerer's Stone  487
The Da Vinci Code                      455
The Twilight Saga                      401
Lord of the Rings                      278
                                     ...
Building Recommendation Engines in Python

Finding the most popular items

print(book_df.value_counts().index)

Index(['40 Shades of Grey', 'Harry Potter and the Sorcerer's Stone',
       'The Da Vinci Code', 'The Twilight Saga',
       'The Lord of the Rings'],
      dtype='object')
Building Recommendation Engines in Python

Finding the most liked items

user_ratings DataFrame:

User Book Rating
User_233 The Great Gatsby 3.0
User_651 The Catcher in the Rye 5.0
User_131 The Lord of the Rings 3.0
User_965 Little Women 4.0
User_651 Fifty Shades of Grey 2.0
... ... ...
Building Recommendation Engines in Python

Finding the most liked items

avg_rating_df = user_ratings[["book", "rating"]].groupby(['book']).mean()
avg_rating_df.head()

                                      rating
title                                      
Hamlet                                   4.1
The Da Vinci Code                       2.1
Gone with the Wind                       4.2
Fifty Shades of Grey                     1.2
Wuthering Heights                        3.9
                                          ...
Building Recommendation Engines in Python

Finding the most liked items

sorted_avg_rating_df = avg_rating_df.sort_values(by="rating", ascending=False)
sorted_avg_rating_df.head()

                                      rating
title                                      
The Girl in the Fog                      5.0
Behind the Bell                          5.0
Across the River and into the Trees      5.0
The Complete McGonagall                  5.0
What Is to Be Done?                      5.0
                                          ...
Building Recommendation Engines in Python

Finding the most liked items

(user_ratings['title']=='The Girl in the Fog').sum()

1

(user_ratings['title']=='Valley of the Dolls').sum()

1

(user_ratings['title']=='Across the River and into the Trees').sum()

1
Building Recommendation Engines in Python

Finding the most liked popular items

book_frequency = user_ratings["book"].value_counts()
print(book_frequency)

40 Shades of Grey                      524
Harry Potter and the Sorcerer's Stone  487
                                       ...

frequently_reviewed_books = book_frequency[book_frequency > 100].index
print(frequently_reviewed_books)

Index([u'The Lord of the Rings', u'To Kill a Mockingbird', u'Of Mice and Men',
       u'1984', u'Hamlet'])
Building Recommendation Engines in Python

Finding the most liked popular items

frequent_books_df =  user_ratings_df[user_ratings_df["book"].isin(frequently_reviewed_books)]

frequent_books_avgs = frequently_reviewed_books[["title", "rating"]].groupby('title').mean()
print(frequent_books_avgs.sort_values(by="rating", ascending=False).head())

                                      rating
title                                      
To Kill a Mockingbird                    4.7
1984.                                    4.7
Harry Potter and the Sorcerer's Stone    4.6
                                          ...
Building Recommendation Engines in Python

Let's practice!

Building Recommendation Engines in Python

Preparing Video For Download...