Page Recommendations - Problem

Imagine you're building a social media recommendation system similar to Facebook or Instagram! You have two important pieces of data:

  • Friendship connections between users
  • Page likes from different users

Your task is to create a smart recommendation engine that suggests pages to user_id = 1 based on what their friends have liked. The catch? Don't recommend pages they've already liked themselves!

Rules:

  • Find all friends of user 1
  • Collect all pages that these friends have liked
  • Filter out pages that user 1 already likes
  • Return unique page recommendations

This is a classic collaborative filtering problem that powers recommendation systems across the web!

Input & Output

example_1.sql — Basic Recommendation
$ Input: Friendship: +----------+----------+ | user1_id | user2_id | +----------+----------+ | 1 | 2 | | 1 | 3 | | 1 | 4 | +----------+----------+ Likes: +---------+---------+ | user_id | page_id | +---------+---------+ | 1 | 88 | | 2 | 23 | | 3 | 24 | | 4 | 56 | | 3 | 33 | | 3 | 77 | +---------+---------+
› Output: +---------+ | page_id | +---------+ | 23 | | 24 | | 33 | | 56 | | 77 | +---------+
💡 Note: User 1 is friends with users 2, 3, and 4. These friends like pages [23, 24, 56, 33, 77]. User 1 already likes page 88, which is not in the friends' list, so all friend-liked pages are recommended.
example_2.sql — With Overlapping Likes
$ Input: Friendship: +----------+----------+ | user1_id | user2_id | +----------+----------+ | 1 | 2 | | 3 | 1 | +----------+----------+ Likes: +---------+---------+ | user_id | page_id | +---------+---------+ | 1 | 11 | | 1 | 12 | | 2 | 11 | | 2 | 13 | | 3 | 12 | | 3 | 14 | +---------+---------+
› Output: +---------+ | page_id | +---------+ | 13 | | 14 | +---------+
💡 Note: User 1 is friends with users 2 and 3. Friends like pages [11, 13, 12, 14]. User 1 already likes pages 11 and 12, so only pages 13 and 14 are recommended.
example_3.sql — No Recommendations
$ Input: Friendship: +----------+----------+ | user1_id | user2_id | +----------+----------+ | 1 | 2 | +----------+----------+ Likes: +---------+---------+ | user_id | page_id | +---------+---------+ | 1 | 10 | | 1 | 20 | | 2 | 10 | | 2 | 20 | +---------+---------+
› Output: +---------+ | page_id | +---------+ (empty result)
💡 Note: User 1 is friends with user 2. Friend likes pages [10, 20]. User 1 already likes both pages 10 and 20, so no new recommendations are available.

Constraints

  • 1 ≤ user_id ≤ 104
  • 1 ≤ page_id ≤ 104
  • There are no duplicate rows in the Friendship table
  • There are no duplicate rows in the Likes table
  • Friendship is bidirectional - if (A,B) exists, then A and B are friends
  • User 1 is guaranteed to exist in the database

Visualization

Tap to expand
User 1Friend 2Friend 3Friend 4Page 23Page 24Page 33Page 56Page 77Recommendations23, 24, 33, 56, 77Already likes: 88Filter out existing likesPage Recommendation SystemCollaborative Filtering Based on Friend Activity
Understanding the Visualization
1
Find Your Circle
Identify all friends of user 1 (bidirectional relationship)
2
Gather Preferences
Collect all pages that these friends have liked
3
Remove Duplicates
Filter out pages that user 1 has already liked
4
Generate Recommendations
Return the unique set of recommended pages
Key Takeaway
🎯 Key Insight: Use efficient set operations (UNION for friends, EXCEPT for filtering) to minimize database scans and achieve optimal O(n) performance in recommendation systems.

Related Problems

Asked in
Meta 45 LinkedIn 38 Twitter 32 Instagram 28
42.9K Views
High Frequency
~15 min Avg. Time
1.8K Likes
Ln 1, Col 1
Smart Actions
💡 Explanation
AI Ready
💡 Suggestion Tab to accept Esc to dismiss
// Output will appear here after running code
Code Editor Closed
Click the red button to reopen