• 한국경영과학회:학술대회논문집
• /
• 대한산업공학회/한국경영과학회 2004년도 춘계공동학술대회 논문집
• /
• pp.702-705
• /
• 2004

Collaborative filtering (CF) is originally based on the ratings of customers who vote on the items they used. When customers' votes are not available, user-item binary data set which represents choice and non-choice can also be used in this analysis. In this case the similarities between active user and the other users must be modified. Therefore we compare eight types of binary similarities by applying them in the modified CF Algorithm. Some experimental results will be reported.

• Journal of Information Science Theory and Practice
• /
• 제2권3호
• /
• pp.18-28
• /
• 2014

Reddit is a large user generated content (USG) website in which users form common interest groups and submit links to external content or text posts of user-created content. The web site operates on a voting system whereby registered users can assign positive or negative ratings to both submitted content and comments made to submitted content. While Reddit is a pseudonymous site, with users creating usernames but providing no biographical data, an informal survey posted to a large shared interest community yielded 734 responses including age and gender of users. This provided a large amount of contextual biographical data with which to analyse user profiles at the first level of Computer Mediated Discourse Analysis (CMDA), articulated by Susan Herring. The results indicate that older Reddit users both formulate more complex writing and enjoy more success when rated by other users. Gender data was incomplete and as such only tentative results could be proposed in that regard.

• Journal of Information Processing Systems
• /
• 제15권1호
• /
• pp.127-136
• /
• 2019

User-based and item-based approaches have been developed as the solutions of the movie recommendation problem. However, the user-based approach is faced with the problem of sparsity, and the item-based approach is faced with the problem of not reflecting users' preferences. In order to solve these problems, there is a research on the combination of the two methods using the concept of similarity. In reality, it is not free from the problem of sparsity, since it has a lot of parameters to be calculated. In this study, we propose a combining method that simplifies the combination equation of prior study. This method is relatively free from the problem of sparsity, since it has less parameters to be calculated. Thus, it can get more accurate results by reflecting the users rating to calculate the parameters. It is very fast to predict new movie ratings as well. In experiments for the proposed method, the initial error is large, but the performance gets quickly stabilized after. In addition, it showed about 6% lower average error rate than the existing method using similarity.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제31권8호
• /
• pp.1062-1072
• /
• 2004

협력적 여과 시스템은 ｛사용자-문서｝의 행렬을 기반으로 사용자에게 웹 문서를 추천하는 데 있어서 효율적인 시스템이다. 그러나 협력적 여과 시스템은 초기 평가 문제와 희박성으로 인하여 추천의 정확도가 저하된다는 단점을 갖는다. 본 논문에서는 협력적 여과 시스템의 희박성과 초기 평가 문제를 해결하기 위하여 사용자 프로파일을 생성시킴으로써 자동으로 선호도를 평가하는 방법을 제안한다. 본 논문에서 사용하는 프로파일은 협력적 여과 시스템에서의 ｛사용자-문서｝ 행렬을 기반으로 생성된 사용자 프로파일에 내용 기반 여과 시스템에서 연관 피드백을 이용하여 생성한 사용자 프로파일을 상호정보의 방법에 의해 병합함으로써 생성한 내용 기반 협력적 사용자 프로파일이다. 생성한 내용 기반 협력적 사용자 프로파일을 정규화시키고, 정규화한 프로파일을 협력적 여과 시스템의 ｛사용자-문서｝ 행렬에 반영함으로써 자동으로 선호도를 평가한다. 제안된 방법은 사용자가 웹 문서에 대해서 선호도를 평가한 데이터베이스에서 평가되었으며, 기존의 방법보다 보다 효율적임을 증명한다.

• Journal of Information Science Theory and Practice
• /
• 제2권3호
• /
• pp.6-17
• /
• 2014

The purpose of this study is to examine to what extent the context in which people interact with online information affects people's credibility perceptions. In this study, credibility assessment is defined as perceptions of credibility relying on individuals' expertise and knowledge. Context has been characterized with respect to three aspects: Context as user goals and intentions, context as topicality of information, and context as information activities. The data were collected from two empirical studies. Study 1 was a diary study in which 333 residents in Michigan, U.S.A. submitted 2,471 diary entries to report their trust perceptions associated with ten different user goals and nine different intentions. Study 2 was a lab-based study in which 64 subjects participated in performing four search tasks in two different information activity conditions - information search or content creation. There are three major findings of this study: (1) Score-based trust perceptions provided limited views of people's credibility perceptions because respondents tended to score trust ratings consistently high across various user goals and intentions; (2) The topicality of information mattered more when study subjects assessed the credibility of user generated content (UGC) than with traditional media content (TMC); (3) Subjects of this study exerted more effort into making credibility judgments when they engaged in searching activities than in content creation. These findings indicate that credibility assessment can or should be seen as a process-oriented notion incorporating various information use contexts beyond simple rating-based evaluation. The theoretical contributions for information scientists and practical implications for web designers are also discussed.

• 한국인터넷방송통신학회논문지
• /
• 제23권5호
• /
• pp.1-8
• /
• 2023

교차 도메인 추천은 다른 도메인에 있는 관련 사용자 정보 데이터와 아이템 데이터를 공유하는 방법입니다. 주로 사용자 중복이 많은 온라인 쇼핑몰이나 유튜브, 넷플릭스와 같은 멀티미디어 서비스 컨텐츠에서 사용됩니다. K-means 클러스터링을 통해 사용자 데이터와 평점을 기반으로 군집화를 실시하여 임베딩을 생성합니다. 이 결과를 트랜스포머 네트워크를 통해 학습한 후 사용자 만족도를 예측합니다. 그런 다음 트랜스포머 기반 추천 모델을 사용하여 사용자에게 적합한 아이템을 추천합니다. 이 연구를 통해 추천함으로써 더 적은 시간적 비용으로 초기 사용자 문제를 예측하고 사용자들의 만족도를 높일 수 있다는 결과를 실험을 통해 보여주었습니다.

• 지능정보연구
• /
• 제21권2호
• /
• pp.1-18
• /
• 2015

협업 필터링은 학계나 산업계에서 우수한 성능으로 인해 많이 사용되는 추천기법이지만, 정량적 정보인 사용자들의 평가점수에만 국한하여 추천결과를 생성하므로 간혹 정확도가 떨어지는 문제가 발생한다. 이에 새로운 정보를 추가로 고려하여, 협업 필터링의 성능을 개선하려는 연구들이 지금까지 다양하게 시도되어 왔다. 본 연구는 최근 Web 2.0 시대의 도래로 인해 사용자들이 구입한 상품에 대한 솔직한 의견을 인터넷 상에 자유롭게 표현한다는 점에 착안하여, 사용자가 직접 작성한 리뷰를 참고하여 협업 필터링의 성능을 개선하는 새로운 추천 알고리즘을 제안하고, 이를 스마트폰 앱 추천 시스템에 적용하였다. 정성 정보인 사용자 리뷰를 정량화하기 위해 본 연구에서는 텍스트 마이닝을 활용하였다. 구체적으로 본 연구의 추천시스템은 사용자간 유사도를 산출할 때, 사용자 리뷰의 유사도를 추가로 반영하여 보다 정밀하게 사용자간 유사도를 산출할 수 있도록 하였다. 이 때, 사용자 리뷰의 유사도를 산출하는 접근법으로 중복 사용된 색인어의 빈도로 산출하는 방안과 TF-IDF(Term Frequency - Inverse Document Frequency) 가중치 합으로 산출하는 2가지 방안을 제시한 뒤 그 성능을 비교해 보았다. 실험결과, 제안 알고리즘을 통한 추천, 즉 사용자 리뷰의 유사도를 추가로 반영하는 알고리즘이 평점만을 고려하는 전통적인 협업 필터링과 비교해 더 우수한 예측정확도를 나타냄을 확인할 수 있었다. 아울러, 중복 사용 단어의 TF-IDF 가중치의 합을 고려했을 때, 단순히 중복 사용 단어의 빈도만을 고려했을 때 보다 조금 더 나은 예측정확도를 얻을 수 있음도 함께 확인할 수 있었다.

• 한국철도학회:학술대회논문집
• /
• 한국철도학회 2008년도 춘계학술대회 논문집
• /
• pp.1853-1859
• /
• 2008

IGBT modules are designed for low loss, rugged for all environments and user friendly. Low on state saturation voltage with high switching speed is the primary concerns. In this paper selection of IGBT, module ratings and characteristics are discussed. The IGBT design topic of protection against over voltage and over current are covered. Emphasis on turn off switching, short circuit switching and necessary precautions are dealt. Selection of IGBT device, gate drive power, and its lay out considerations are covered in detail.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2019년도 추계학술발표대회
• /
• pp.716-719
• /
• 2019

Collaborative filtering has been used frequently as a recommendation system. To reduce the errors on predicting the ratings that may be given by the user, we propose a new aggregation method to do so. We used a real-world dataset MovieLens to compare our proposed method from previously existing methods, and accordingly to the results, ours was more accurate.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권12호
• /
• pp.3798-3814
• /
• 2022

Social recommendation algorithm can alleviate data sparsity and cold start problems in recommendation system by integrated social information. Among them, matrix-based decomposition algorithms are the most widely used and studied. Such algorithms use dot product operations to calculate the similarity between users and items, which ignores user's potential preferences, reduces algorithms' recommendation accuracy. This deficiency can be avoided by a metric learning-based social recommendation algorithm, which learns the distance between user embedding vectors and item embedding vectors instead of vector dot-product operations. However, previous works provide no theoretical explanation for its plausibility. Moreover, most works focus on the indirect impact of social friends on user's preferences, ignoring the direct impact on user's rating preferences, which is the influence of user rating preferences. To solve these problems, this study proposes a user bias drift social recommendation algorithm based on metric learning (BDML). The main work of this paper is as follows: (1) the process of introducing metric learning in the social recommendation scenario is introduced in the form of equations, and explained the reason why metric learning can replace the click operation; (2) a new user bias is constructed to simultaneously model the impact of social relationships on user's ratings preferences and user's preferences; Experimental results on two datasets show that the BDML algorithm proposed in this study has better recommendation accuracy compared with other comparison algorithms, and will be able to guarantee the recommendation effect in a more sparse dataset.