• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2005년도 춘계학술대회
• /
• pp.232-239
• /
• 2005

Collaborative-filtering-enabled Web sites that recommend books, CDs, movies, and so on, have become very popular on the Internet. Such sites recommend items to a user on the basis of the opinions of other users with similar tastes. This paper discuss an approach to collaborative filtering based on the Simple Bayesian and apply this model to two variants of the collaborative filtering. One is user-based collaborative filtering, which makes predictions based on the users' similarities. The other is item-based collaborative filtering which makes predictions based on the items' similarities. To evaluate the proposed algorithms, this paper used a database of movie recommendations. Empirical results show that the proposed Bayesian approaches outperform typical correlation-based collaborative filtering algorithms.

• 한국정보처리학회논문지
• /
• 제7권5S호
• /
• pp.1599-1608
• /
• 2000

Automated collaborative filtering is on the verge of becoming a popular technique to reduce overloaded information as well as to solve the problems that content-based information filtering systems cannot handle. In this paper, we describe three different algorithms that perform collaborative filtering: GroupLens that is th traditional technique; Best N, the modified one; and an algorithm that uses clustering. Based on the exeprimental results using real data, the algorithm using clustering is compared with the existing representative collaborative filtering agent algorithms such as GroupLens and Best N. The experimental results indicate that the algorithms using clustering is similar to Best N and better than GroupLens for prediction accuracy. The results also demonstrate that the algorithm using clustering produces the best performance according to the standard deviation of error rate. This means that the algorithm using clustering gives the most stable and the best uniform recommendation. In addition, the algorithm using clustering reduces the time of recommendation.

• Journal of information and communication convergence engineering
• /
• 제17권2호
• /
• pp.135-141
• /
• 2019

Collaborative filtering algorithms often encounter data sparsity issues. To overcome this issue, auxiliary information of relevant items is analyzed and an item attribute matrix is derived. In this study, we combine the user-item attribute preference with the traditional similarity calculation method to develop an improved similarity calculation approach and use weights to control the importance of these two elements. A collaborative filtering algorithm based on user-item attribute preference is proposed. The experimental results show that the performance of the recommender system is the most optimal when the weight of traditional similarity is equal to that of user-item attribute preference similarity. Although the rating-matrix is sparse, better recommendation results can be obtained by adding a suitable proportion of user-item attribute preference similarity. Moreover, the mean absolute error of the proposed approach is less than that of two traditional collaborative filtering algorithms.

• 한국인터넷방송통신학회논문지
• /
• 제10권3호
• /
• pp.51-59
• /
• 2010

Collaborative Filtering(CF) 기법에 기반을 둔 다양한 영화 추천 방법들이 제안 되어 왔다. CF는 영화를 본 사람들이 직접 영화에 대해 평가한 점수를 기반으로 같은 성향을 가진 이웃 그룹을 결정하고, 새로운 영화에 대해 그 영화를 이미 본 이웃의 점수를 기반으로 사용자의 새로운 영화에 대한 선호도 값을 예측하는 방법이다. 본 논문에서는 사용자에 따라 영화 장르에 대한 선호도 정보를 CF의 예측 값에 반영하는 새로운 방법을 제안한다. 이 방법은 CF를 기반으로 하는 모든 종류의 추천 방법에 결합하여 사용할 수 있다. 본 논문에서는 기존의 CF알고리즘에 장르기반 알고리즘을 결합한 CF-Genre의 성능과 기존의 CF 알고리즘의 성능을 측정 비교하였다. 성능 평가의 결과 CF-Genre가 기존 CF의 예측 성능을 3.3% 정도 개선하였다.

• 창의정보문화연구
• /
• 제4권3호
• /
• pp.221-230
• /
• 2018

추천 시스템의 주요 방법인 협력 필터링 기술은 실제 상업용 온라인 시스템에서 성공적으로 구현되어 서비스가 제공되고 있다. 그러나, 이 기술은 본질적으로 여러 가지 단점을 내포하는데, 데이터 희소성, 콜드 스타트, 확장성 문제 등이 그 예이다. 확장성 문제를 해결하기 위하여 클러스터링 기법을 활용한 협력 필터링 방법이 연구되어 왔다. 본 연구에서 제안하는 협력 필터링 시스템에서는 가장 널리 활용되는 클러스터링 기법들 중 하나인 K-means 알고리즘의 단점을 개선하고자 유전자 알고리즘을 이용한다. 또한, 기존 연구에서 최적화된 클러스터링 결과를 추구하였던 것과는 달리, 제안 방법은 클러스터링 결과를 활용한 협력 필터링 시스템 성능의 최적화를 목표로 하므로, 실질적으로 시스템의 성능을 향상시킬 수 있다.

• Journal of information and communication convergence engineering
• /
• 제22권1호
• /
• pp.7-13
• /
• 2024

User preferences and ratings may be anticipated by recommendation systems, which are widely used in social networking, online shopping, healthcare, and even energy efficiency. Constructing trustworthy recommender systems for various applications, requires the analysis and mining of vast quantities of user data, including demographics. This study focuses on holding elections with vague voter and candidate preferences. Collaborative user ratings are used by filtering algorithms to provide suggestions. To avoid information overload, consumers are directed towards items that they are more likely to prefer based on the profile data used by recommender systems. Better interactions between governments, residents, and businesses may result from studies on recommender systems that facilitate the use of e-government services. To broaden people's access to the democratic process, the concept of "e-democracy" applies new media technologies. This study provides a framework for an electronic voting advisory system that uses machine learning.

• Management Science and Financial Engineering
• /
• 제12권2호
• /
• pp.71-85
• /
• 2006

Collaborative filtering, among other recommender systems, has been known as the most successful recommendation technique. However, it requires the user-item rating data, which may not be easily available. As an alternative, some collaborative filtering algorithms have been developed recently by utilizing the market basket data in the form of the binary user-item matrix. Viewing the recommendation scheme as a two-class classification problem, we proposed a new collaborative filtering scheme using a regularized discriminant analysis applied to the binary user-item data. The proposed discriminant model was built in terms of the major principal components and was used for predicting the probability of purchasing a particular item by an active user. The proposed scheme was illustrated with two modified real data sets and its performance was compared with the existing user-based approach in terms of the recommendation precision.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제7권5호
• /
• pp.1272-1290
• /
• 2013

Privacy-preserving collaborative filtering schemes are becoming increasingly popular because they handle the information overload problem without jeopardizing privacy. However, they may be susceptible to shilling or profile injection attacks, similar to traditional recommender systems without privacy measures. Although researchers have proposed various privacy-preserving recommendation frameworks, it has not been shown that such schemes are resistant to profile injection attacks. In this study, we investigate two memory-based privacy-preserving collaborative filtering algorithms and analyze their robustness against several shilling attack strategies. We first design and apply formerly proposed shilling attack techniques to privately collected databases. We analyze their effectiveness in manipulating predicted recommendations by experimenting on real data-based benchmark data sets. We show that it is still possible to manipulate the predictions significantly on databases consisting of masked preferences even though a few of the attack strategies are not effective in a privacy-preserving environment.

• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2003년도 춘계학술대회
• /
• pp.400-409
• /
• 2003

Collaborative filtering (CF) recommendation is a knowledge sharing technology for distribution of opinions and facilitating contacts in network society between people with similar interests. The main concerns of the CF algorithm are about prediction accuracy, speed of response time, problem of data sparsity, and scalability. In general, the efforts of improving prediction algorithms and lessening response time are decoupled. We propose a three-step CF recommendation model which is composed of profiling, inferring, and predicting steps while considering prediction accuracy and computing speed simultaneously. This model combines a CF algorithm with two machine learning processes, SOM (Self-Organizing Map) and CBR (Case Based Reasoning) by changing an unsupervised clustering problem into a supervised user preference reasoning problem, which is a novel approach for the CF recommendation field. This paper demonstrates the utility of the CF recommendation based on SOM cluster-indexing CBR with validation against control algorithms through an open dataset of user preference.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권5호
• /
• pp.1341-1368
• /
• 2024

The concept of privacy-preserving collaborative filtering (PPCF) has been gaining significant attention. Due to the fact that model-based recommendation methods with privacy are more efficient online, privacy-preserving memory-based scheme should be avoided in favor of model-based recommendation methods with privacy. Several studies in the current literature have examined ant colony clustering algorithms that are based on non-privacy collaborative filtering schemes. Nevertheless, the literature does not contain any studies that consider privacy in the context of ant colony clustering-based CF schema. This study employed the ant colony clustering model-based PPCF scheme. Attacks like shilling or profile injection could potentially be successful against privacy-preserving model-based collaborative filtering techniques. Afterwards, the scheme's robustness was assessed by conducting a shilling attack using six different attack models. We utilize masked data-based profile injection attacks against a privacy-preserving ant colony clustering-based prediction algorithm. Subsequently, we conduct extensive experiments utilizing authentic data to assess its robustness against profile injection attacks. In addition, we evaluate the resilience of the ant colony clustering model-based PPCF against shilling attacks by comparing it to established PPCF memory and model-based prediction techniques. The empirical findings indicate that push attack models exerted a substantial influence on the predictions, whereas nuke attack models demonstrated limited efficacy.