• Journal of the Korean Data and Information Science Society
• /
• 제18권3호
• /
• pp.619-628
• /
• 2007

This study is devoted to suggesting the norm of detection abnormal users who are inferior to the other users in the recommender system compared with estimation accuracy. To select the abnormal users, we propose the pre-filtering method by using the preference ratings to the item rated by users. In this study, the experimental result shows the possibility of detecting the abnormal users before the process of preference estimation through the prediction algorithm. And It will be possible to improve the performance of the recommender system by using this detecting norm.

• 정보처리학회논문지D
• /
• 제9D권2호
• /
• pp.329-336
• /
• 2002

인터넷 환경의 급속한 발전과 함께 이를 이용한 전자상거래가 빠르게 증가하고 있다. 증가하는 전자상거래 환경에서 고객에게 필요한 제품을 신속히 제공하고, 제품판매를 증가시킬 수 있는 새로운 전자상거래 시스템의 필요성이 점차 커지고 있다. 이러한 필요성에 의해서 최근에 추천시스템에 대한 많은 연구가 이루어지고 있다. 하지만 지금까지의 추천시스템은 고객의 구매데이터가 증가하면 고객에게 추천을 제공하는데 많은 시간이 소요되어 실시간 추천이 어렵다는 큰 단점을 가졌다. 따라서, 이 논문은 전자상거래 시스템의 경쟁력을 높이는 방안으로 협동적 필터링을 이용한 추천시스템을 연구하고, 성능을 개선하기 위해서 추천에 사용되는 데이터를 제품의 대표장르를 이용하여 줄임으로서 추천소요시간을 단축하여 실시간 추천이 가능한 개선된 추천시스템을 제안하고 실험하였다. 또한 개선된 추천시스템을 Enterprise JavaBeans로 구현함으로서 분산환경에서 사용할 수 있는 전자상거래시스템을 설계하여 경쟁력있는 전자상거래 시스템 환경을 제공하고자 한다.

• 융합정보논문지
• /
• 제9권6호
• /
• pp.83-90
• /
• 2019

사람들은 자신의 더 나은 선택을 위하여 끊임없이 노력한다. 이러한 이유로 추천시스템이 개발되었으며, 1990년대 초반부터 계속해서 발전하고 있다. 그 중, 협업필터링 기법은 추천시스템 분야에서 우수한 성능을 보였으며, 기계학습이 등장하면서 기계학습을 이용한 추천시스템에 관한 연구가 활발히 진행되었다. 본 연구는 앙상블 방법 중에서 스태킹 모형을 사용하여 추천시스템을 구축하며, 실제 고객의 상품 구매 데이터를 활용하여 협업필터링과 기계학습 기반 스태킹 모형으로 추천시스템을 개발하였다. 제시한 모형의 추천 성능은 기존의 협업필터링과 기계학습 기반 추천시스템과 비교하여 모형의 우수성을 확인하며, 연구결과는 스태킹 모형을 이용한 추천시스템 모형의 추천 성능이 개선됨을 확인하였다. 향후 본 연구에서 제안한 모형은 개인이나 기업이 더 나은 선택을 하여 상품을 추천할 때 도움을 줄 것으로 기대한다.

• 대한산업공학회지
• /
• 제41권2호
• /
• pp.185-208
• /
• 2015

The explosive growth of the world-wide-web and the emergence of e-commerce has led to the development of recommender systems. Recommender systems are personalized information filtering used to identify a set of items that will be of interest to a certain user. This paper reviews recommender systems and presents their pros and cons.

• 대한안전경영과학회지
• /
• 제12권1호
• /
• pp.127-136
• /
• 2010

In this paper, the author proposed following two methods to improve the accuracy of the recommender system. First, in order to classify the users more accurately, the author used a EMC(Expanded Moving Center) heuristic algorithm which improved clustering accuracy. Second, the author proposed the Neighborhood-oriented preference prediction method that improved the conventional preference prediction methods, so the accuracy of the recommender system is improved. The test result of the recommender system which adapted the above two methods suggested in this paper was improved the accuracy than the conventional recommendation methods.

• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2005년도 춘계학술대회
• /
• pp.131-138
• /
• 2005

Recommender systems have been widely advocated as a way of coping with the problem of information overload in e-business environment. Most of the existing recommender systems focused on what kind of items to recommend, although when to recommend to the target customer considering their context is an important issue. Even right item might be a spam advertisement or wrong recommendation for the customer if it can not be recommended at the right context. It is particularly important for recommendations where the user's context is changing rapidly, such as in both handheld and ubiquitous computing environment. Therefore, we propose CARS (Context-Aware Recommender System) based on CBR and context-awareness for ubiquitous computing environment. CBR is used to generate a target customer class and proper context. Context-awareness is used to gather suer context information from sensors, networks, device status, user profiles, and other sources. An illustrative case example is suggested to explain the procedure of CARS.

• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2005년도 공동추계학술대회
• /
• pp.415-424
• /
• 2005

The rapid growth of blog has caused information overload where bloggers in the virtual community space are no longer able to effectively choose the blogs they are exposed to. Recommender systems have been widely advocated as a way of coping with the problem of information overload in e-business environment. Collaborative Filtering (CF) is the most successful recommendation method to date and used in many of the recommender systems. Therefore, we propose a CF-based recommender system for bloggers in the virtual community space. Our proposed methodology consists of three main phases: In the first phase, we apply the "Interest Value" to a recommender system. The Interest Value is a quantity value about user preference in virtual community, and can measure the opinion of users accurately. Next phase, we generate the neighborhood group based on the Interest Value. In the final phase, we use the Community Likeness Score (CLS) to generate the top-n recommendation list. The methodology is explained step by step with an illustrative example and is verified with real data of a blog service provider.

• 한국산학기술학회:학술대회논문집
• /
• 한국산학기술학회 2003년도 Proceeding
• /
• pp.113-117
• /
• 2003

As the wireless communication technology advances rapidly, a personalization technology can be incorporated with the mobile Internet environment, which is based on location-based services to support more accurate personalized services. A location-based personalized recommender system is one of the essential technologies of the location-based application services, and is also a crucial technology for the ubiquitous environment. In this paper we propose a framework of a location-based personalized recommender system for the mobile Internet environment. The proposed system consists of three modules the interface module, the neighbor selection module and the prediction and recommendation module. The proposed system incorporates the concept of the recommendation system in the Electronic Commerce along with that of the mobile devices for possible expansion of services on the mobile devices. Finally a service scenario for entertainment recommendation based on the proposed recommender system is described.

• 제어로봇시스템학회논문지
• /
• 제17권8호
• /
• pp.739-746
• /
• 2011

Recently, recommender systems have been widely applied in E-commerce websites to help their customers find the items what they want. A recommender system should be able to provide users with useful information regarding their interests. The ability to immediately respond to the changes in user's preference is a valuable asset of recommender systems. This paper proposes a novel recommender system which aims to effectively adapt and respond to the immediate changes in user's preference. The proposed system combines IEC (Interactive Evolutionary Computation) with a content-based filtering method and also employs data grouping in order to improve time efficiency. Experiments show that the proposed system makes acceptable recommendations while ensuring quality and speed. From a comparative experimental study with an existing recommender system which uses the content-based filtering, it is revealed that the proposed system produces more reliable recommendations and adaptively responds to the changes in any given condition. It denotes that the proposed approach can be an alternative to resolve limitations (e.g., over-specialization and sparse problems) of the existing methods.

• 대한산업공학회지
• /
• 제36권3호
• /
• pp.193-202
• /
• 2010

The recommender system is a typical software solution for personalized services which are now popular in e-commerce sites. Most of the existing recommender systems are based on customers' explicit rating data on items (e.g., ratings on movies), and it is only recently that recommender systems based on implicit ratings have been proposed as a better alternative. Implicit ratings of a customer on those items that are clicked but not purchased can be inferred from the customer's navigational and behavioral patterns. In this article, a dimensionality reduction (DR) technique is newly applied to the implicit rating-based recommender system, and its effectiveness is assessed using an experimental e-commerce site. The experimental results indicate that the performance of the proposed approach is superior or at least similar to the conventional collaborative filtering (CF)-based approach unless the number of recommended products is 'large.' In addition, the proposed approach requires less memory space and is computationally more efficient.