블로그 이미지
가요음반 (CD/LP) 수집가 ......................... / maniadb.com (불량)운영자 ................. / social music service 24hz.com / 트위터 : xfactor_ryu
xfactor

Recent Comment

Recent Trackback

Archive

calendar

          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29  
  • 511,573total
  • 76today
  • 101yesterday
2008/02/01 14:00 study/news
음원 저작권 침해 문제에서 정부는 빠지라는 주장이 RIAA(전미 저작권 협회)서 나왔다. RIAA 캐리 셔먼은 저작권 침해를 혐오, 그 벌칙을 강화하는 법률은 찬성하지만 정부의 개입은 반대한다는 뜻을 밝혔다.

셔먼은 30일(미국시간) ‘State of the Net’ 컨퍼런스에서 “ISP(인터넷 서비스 프로바이더)'를 정부가 법으로 압박하는 것보다 시장에 맞는 방법을 찾아내는 것이 더 중요하다”고 밝혔다.

셔먼이 말하는 ‘법의 압박’은 1998년 제정된 디지털 밀레니엄 저작권법(Digital Millennium Copyright Act:DMCA)을 뜻한다. 이 법률에는 ISP가 불법 콘텐츠를 적발당한 시점에서 그 공개를 중지한다면 제제를 가하지 않는다는 조항이 있다.

...
http://www.zdnet.co.kr/news/digital/0,39030978,39165630,00.htm
크리에이티브 커먼즈 라이선스
Creative Commons License
posted by xfactor
2008/02/01 13:45 study/research

Item-based Collaborative Filtering Recommendation Algorithms

Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl
{sarwar, karypis, konstan, riedl}@cs.umn.edu
GroupLens Research Group/Army HPC Research Center
Department of Computer Science and Engineering
University of Minnesota, Minneapolis, MN 55455

Copyright is held by the author/owner(s).

WWW10, May 1-5, 2001, Hong Kong.

ACM 1-58113-348-0/01/0005.

Abstract:

Recommender systems apply knowledge discovery techniques to the problem of making personalized recommendations for information, products or services during a live interaction. These systems, especially the k-nearest neighbor collaborative filtering based ones, are achieving widespread success on the Web. The tremendous growth in the amount of available information and the number of visitors to Web sites in recent years poses some key challenges for recommender systems. These are: producing high quality recommendations, performing many recommendations per second for millions of users and items and achieving high coverage in the face of data sparsity. In traditional collaborative filtering systems the amount of work increases with the number of participants in the system. New recommender system technologies are needed that can quickly produce high quality recommendations, even for very large-scale problems. To address these issues we have explored item-based collaborative filtering techniques. Item-based techniques first analyze the user-item matrix to identify relationships between different items, and then use these relationships to indirectly compute recommendations for users.

In this paper we analyze different item-based recommendation generation algorithms. We look into different techniques for computing item-item similarities (e.g., item-item correlation vs. cosine similarities between item vectors) and different techniques for obtaining recommendations from them (e.g., weighted sum vs. regression model). Finally, we experimentally evaluate our results and compare them to the basic k-nearest neighbor approach. Our experiments suggest that item-based algorithms provide dramatically better performance than user-based algorithms, while at the same time providing better quality than the best available user-based algorithms.



* 출처 : http://www10.org/cdrom/papers/519
크리에이티브 커먼즈 라이선스
Creative Commons License
posted by xfactor
2008/02/01 13:21 study/research

정의: 고객들의 선호도와 관심 표현을 바탕으로 선호도 관심에서 비슷한 패턴을 가진 고객들을 식별해 내는 기법 비슷한 취향을 가진 고객들에게 서로 아직 구매하지 않은 상품들은 교차 추천하거나 분류된 고객의 취향이나 생활 형태에 따라 관련 상품을 추천하는 형태의 서비스를 제공하기 위해 사용된다.

협업적 필터링(Collaborative filtering)은 Resnick, P., Iacovou, N., Suchak, M., Bergstrom, P., & Riedle, J.가 1994년에 처음 발표한 개념으로써, 다음의 두 단계를 거친다.

Step1: Neighborhood formation

자신과 가장 유사한 성향을 지닌 사용자 N명을 K-NN 기법에 의해 찾아낸다. 이 때, 각 사용자가 각 item에 대해 rating한 값을 토대로, Pearson correlation을 사용하여 가장 유사한 사용자 그룹(neighbor)을 찾아내는 것이다. 이 때, 목적에 따라 constrained pearson correlation, Cosine vector, Jaccard coefficient, Euclidean distance등이 사용될 수 있으나, Pearson correlation이 가장 보편적으로 사용된다.  (참고: Shardanand, U., & Maes, P.등은 pearson correlation 대신 constrained pearson correlation을 사용하여 예측력을 높였음)

Step2: Recommendation

특정 사용자(user a)와 성향이 유사한 사용자 k명을 찾아낸 후, user a가 rating하지 않은 item에 대한 선호도를 예측한다. 이 때, 앞서 추출된 Neighbor와의 correlation을 가중치로 하여 가중평균합을 구하게 된다. 그 후, 예측값 중 가장 큰 top-N을 추출하여 user a에게 추천하게 된다.

협업적필터링은 매우 간단한 컨셉이다. 예를 들어, 내가 '쉬리','괴물','디워'를 재미있게 보았다고 가정할 때, 나와 영화 선호도가 비슷한 사용자들을 찾아내고, 그 사람들이 만약 '트랜스포머'를 재밌게 봤다고 한다면...바로 '트랜스포머'를 추천하는 것이다.

References

[1] Resnick, P., Iacovou, N., Suchak, M., Bergstrom, P., & Riedle, J. (1994). Grouplens: An open architecture for collaborative filtering of netnews. Proceedings of the ACM 1994 Conference on Computer Supported Cooperative Work (pp.175-186). Chapel Hill.

[2] Shardanand, U., & Maes, P. (1995). Social information filtering: algorithms for automating word of mouth. Proceedings of Conference on Human Factors in Computing Systems (pp.210-217). Denver, CO.



* 출처 : http://blog.naver.com/mrsjk2?Redirect=Log&logNo=60047019060

크리에이티브 커먼즈 라이선스
Creative Commons License
posted by xfactor