Recommender Systems Part 1

1. Introduction
2. Collaborative Filtering – Part 1
3. Programming Work

Transcript

1. Recommender Systems & Collaborative Filtering Yusuke Yamamoto Faculty of Informatics Associate Professor yusuke_yamamoto@acm.org Data Engineering （Recommender System 1） 2022.11.14
2. 1 Introduction to Recommender Systems 2
3. 3 Predicts user preference model and decides which items should be recommended. Recommender System
4. The most familiar recommender system: Amazon(1/2) 4 画像出典：https://www.amazon.co.jp/ Recommend items which users will like
5. The most familiar recommender system: Amazon(2/2) 5 画像出典：https://www.amazon.co.jp/ Recommend items related to a target item
6. SNS x Recommender System 6 User recommendation on Twitter Event recommendation on Facebook
7. Movie recommendation from Netflix 7
8. Apple Music x Recommender System 8
9. News x Recommender System 9
10. Computational Ads 10 画像出典：http://www.apple-style.com
11. Recommendation Systems appear anywhere! 11
12. Why using Recommender Systems? 12 Value for end users • Find interesting items and information easily • Narrow down the set of choices Value for providers • Increase trust and customer loyalty • Increase sales, click rates, conversion rates etc. • Discover new things.. • Opportunities for promotion
13. Paradigms of Recommender System 13 Dietmar Jannach氏のRecommender Systems: An IntroductionのPPT資料より Recommender System item score item1 0.9 item2 1 item3 0.3 … … User profile & context Recommendation list for a target user 1 Community data 2 Item features 3 User preference model
14. 3 main approach for recommendation 14 Collaborative filtering Decides which items should be recommended, based on users behavior logs. Content-based filtering Decides which items should be recommended, based on item features and their metadata Knowledge-based filtering Decides which items should be recommended, based on preference information which users explicitly show
15. Problem Definition 15 § User u’ s behavior data Bu ={b1 , b2 , …, bn } § Item set I = {i1 , i2 , …, im } § User u’s profile（user model）：pu § Relevance between pu and item i ：Rel (pu , i) Input Output Ranked list of item set I （∀ i ∈ I）, based on Rel (pu , i) s.t. l How to model user profiles? l How to compute relevance? Point
16. Content list of this lecture 16 1. Collaborative Filtering (CF) 2. Content-based Filtering 3. Link analysis 4. Advanced CF Lecture + Programming Work so that you can see how methods work
17. References for this lecture 画像出典：https://www.amazon.co.jp/ 17
18. 2 Collaborative Filtering – Part 1 18
19. Collaborative Filtering (CF) 19 Approach Uses user’s rating data to model user preferences and to recommend items Basis assumption • Users give ratings to items appropriately • Patterns in the rating data help us predict the ratings Practical points • Large commercial eCommerce sites use the CF • Well-understood • Applicable in many domains if only rating data can be obtained
20. Example 20 How much does Alice like Item 5？ Q. Alice Item 1 Item 2 Item 3 Item 4 Item 5 ? Items purchased by Alice and her ratings Un-purchased item A ✓ ✓ ✓ ✓
21. Let’s observe other users’ ratings 21 Can we predict Alice’s rating using others’ ratings? Q. Item1 Item2 Item3 Item4 Item5 Alice 5 3 4 4 ? User1 3 1 2 3 3 User2 4 3 4 3 5 User3 3 3 1 5 4 User4 1 5 5 2 1
22. Item1 Item2 Item3 Item4 Item5 Alice 5 3 4 4 ? User1 3 1 2 3 3 User2 4 3 4 3 5 User3 3 3 1 5 4 User4 1 5 5 2 1 22 Dissimilar Similar Let’s observe other users’ ratings Can we predict Alice’s rating using others’ ratings? Q.
23. Item1 Item2 Item3 Item4 Item5 Alice 5 3 4 4 ? User1 3 1 2 3 3 User2 4 3 4 3 5 User3 3 3 1 5 4 User4 1 5 5 2 1 Let’s observe other users’ ratings 23 Alice will give high score to item 5 ! Can we predict Alice’s rating using others’ ratings? Q.
24. User-based Collaborative Filtering 24 Item1 Item2 Item3 Item4 Item5 Alice 5 3 4 4 ? User1 3 1 2 3 3 User2 4 3 4 3 5 User3 3 3 1 5 4 User4 1 5 5 2 1 Focus on ratings of similar users Idea: Point How to compute user similarity How do we combine the ratings of the similar users to predict Alice’s rating? Which/how many similar users should we consider? 1. 2. 3.
25. Similarity between users（1/3） 25 Pearson Correlation Coefficient 𝑠𝑖𝑚 𝑢! , 𝑢" = ∑#∈% (𝑟&!,# − 𝑟&! )(𝑟&",# − 𝑟&" ) ∑ #∈% 𝑟&!,# − 𝑟&! ( ∑ #∈% 𝑟&!,# − 𝑟&! ( ：User a, b 𝑢! , 𝑢" 𝑟#!,% ：User a’s rating to item i 𝐼 : Item set 𝑟#! 𝑟#" , ：User a, b’s average rating
26. Similarity between users（2/3） 26 𝑠𝑖𝑚 𝑢! , 𝑢" = ∑#∈% (𝑟&!,# − 𝑟&! )(𝑟&",# − 𝑟&" ) ∑ #∈% 𝑟&!,# − 𝑟&! ( ∑ #∈% 𝑟&",# − 𝑟&" ( Item1 Item2 Item3 Item4 Alice 5 3 4 4 User1 3 1 2 3 User2 4 3 4 3 User3 3 3 1 5 User4 1 5 5 2 sim=0.85 sim=-0.79 Pearson Correlation Coefficient
27. Similarity between users（3/3） 27 𝑠𝑖𝑚 𝑢! , 𝑢" = ∑#∈% (𝑟&!,# − 𝑟&! )(𝑟&",# − 𝑟&" ) ∑ #∈% 𝑟&!,# − 𝑟&! ( ∑ #∈% 𝑟&",# − 𝑟&" ( Item1 Item2 Item3 Item4 Alice 5 3 4 4 User1 3 1 2 3 User2 4 3 4 3 User3 3 3 1 5 User4 1 5 5 2 sim = ? Let’s calculate Pearson Correlation Coefficient
28. Pearson correlation 28 A measure of the linear correlation between two variables X and Y 0 1 2 3 4 5 6 Item1 Item2 Item3 Item4 Alice User1 User4 Rating score sim(Alice, User4)=-0.79 sim(Alice, User1)=0.85 （It takes differences in rating behavior into account）
29. Predicting rating scores based on user similarity（1/3） 29 A typical prediction function 𝑝𝑟𝑒𝑑𝑖𝑐𝑡 𝑢! , 𝑖 = 𝑟"# + ∑"$∈$$ 𝑠𝑖𝑚(𝑢! , 𝑢% ) 1 (𝑟"$,' − 𝑟"$ ) ∑"$∈$$ 𝑠𝑖𝑚(𝑢! , 𝑢% ) ：Target user a 𝑢& 𝑟#,% ：Rating score of u for item i 𝑖 ：Target item i 𝑟#! ：User a’s average rating score 𝑈' ：A set of similar users to ua
30. Predicting rating scores based on user similarity（2/3） 30 Item5 sim Average rating Alice ? 1 4 User1 3 0.85 2.4 User2 5 0.71 3.8 Similar users 4.0 + 0.85× 3 − 2.4 + 0.71×(5 − 3.8) 0.85 + 0.71 = 4.87 Predicted rating score of Alice for Item5 A typical prediction function 𝑝𝑟𝑒𝑑𝑖𝑐𝑡 𝑢! , 𝑖 = 𝑟"# + ∑"$∈$$ 𝑠𝑖𝑚(𝑢! , 𝑢% ) 1 (𝑟"$,' − 𝑟"$ ) ∑"$∈$$ 𝑠𝑖𝑚(𝑢! , 𝑢% )
31. Predicting rating scores based on user similarity（3/3） 31 Item5 sim Average rating Alice ? 1 4 User1 3 0.85 2.4 User2 5 0.71 3.8 Similar users 4.0 + 0.85× 3 − 2.4 + 0.71×(5 − 3.8) 0.85 + 0.71 = 4.87 Predicted rating score of Alice for Item5 A typical prediction function How to choose similar users? Q. 𝑝𝑟𝑒𝑑𝑖𝑐𝑡 𝑢! , 𝑖 = 𝑟"# + ∑"$∈$$ 𝑠𝑖𝑚(𝑢! , 𝑢% ) 1 (𝑟"$,' − 𝑟"$ ) ∑"$∈$$ 𝑠𝑖𝑚(𝑢! , 𝑢% )
32. How to decide “similar users” (nearest neighbors)? 32 Set a threshold for user similarity • If a user has higher similarity than a threshold, he/she can be regarded as a “similar” user • In worst cases, no similar users will be found Focus on top K similar users （kNN method） • If a user ranks at the top K similarity, he/she can be regarded as a similar user • K is often set to between 50 〜 200 • In worst cases, a system uses rating information of users with low similarity
33. Summary of User-based Collaborative Filtering 33 Basic Approach • User similarities are obtained from a rating matrix • Based on rating scores of similar users, systems predict a rating score of target user for a target item Similarity Calculation Pearson correlation coefficient is often used Selection of Similar Users Top K users with high similarity are often selected as similar users
34. 3 Programming Work 34
35. 35 Visit the following URL: https://recsys2022.hontolab.org/