Recommendation for dummy

2,362 views

Published on

추천시스템 개요 및 분류 등.

Published in: Technology
0 Comments
3 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
2,362
On SlideShare
0
From Embeds
0
Number of Embeds
1,626
Actions
Shares
0
Downloads
23
Comments
0
Likes
3
Embeds 0
No embeds

No notes for slide

Recommendation for dummy

  1. 1. A Brief Introduction to ! Recommendation ! (Fallacies & Understanding) Jeong, Buhwan (Ph.D)
  2. 2. X Data-driven Automated Personalized
  3. 3. Everything, but Nothing
  4. 4. For anyone For one in a group For a person For an item
  5. 5. Explicit Rating vs Implicit Feedback
  6. 6. Content-based Filtering (CBF) Collaborative Filtering (CF)
  7. 7. Model-based CF Memory-based CF Matrix Factorization (MF)
  8. 8. User-orientation vs Item-orientation I Us Me I Is
  9. 9. Similarity Measures ! Many common items between users Many common users between items
  10. 10. Similar Items? Similar Users? MxN Co-occurrence, Set theory, Distance, Correlation, Cosine, Kernel
  11. 11. Hybrid (Ensemble) Explicit Rating Collaborative Filtering User Orientation Implicit Feedback + Content-based Filtering Item Orientation
  12. 12. Search Recommendation Goal Retrieval Discovery Query Keyword User or Item Result Documents Items BM25 CBF PageRank CF Ranking Recency, Quality, Filtering, Diversification
  13. 13. ShoppingHow ! Item- & memory-based CF with implicit feedback Hybrid with CBF using category, mall, brand info.
  14. 14. Curse of Dimensionality
  15. 15. n axa n axN MxN m = Mxa m
  16. 16. MF = SVD = LSA/LSI
  17. 17. Let’s play music
  18. 18. How to Evaluate?
  19. 19. Accuracy vs User Satisfaction
  20. 20. Fast Iteration >> Good Algorithm
  21. 21. Post Analysis & Review
  22. 22. New Perspective ! Netflix’s micro tagging/genre Amazon’s anticipatory shipping
  23. 23. Cold-start Data sparsity Dimensional complexity Coverage Serendipity & Diversity Explainability
  24. 24. PR = P + M + R + F
  25. 25. Just do it.

×