연구실에서 진행 또는 완료된 프로젝트 소개    
A study on attack patterns and attack detection for robust recommender systems
(견고한 추천 시스템을 위한 공격 패턴 및 공격 탐지에 관한 연구)
Funding Korea Research Foundafion
Term 2007. 12 - 2008. 11
P.I Juntae Kim
본 과제의 연구 목표는 협동적 여과 추천 시스템에 대한 악의적 공격 방법 을 분석하고 공격에 대한 탐지 방안을 연구하여 궁극적으로 견고한 추천 시스템을 구현하는 방법을 개발하고 그 효율성을 증명한는 것이다. 이러한 목표를 달성하기 위하여 본 과제에서는 여러 추천 방식의 특징들을 분석하 고, 다양한 공격 방식에 대하여 연구하며, 그동안 연구되어온 공격 탐지 기법들에 대해 그 효과를 분석한다.
Development of Personalized Music Recommendation System
(개인화된 디지털 음악 추천 시스템 개발)
Funding Seoul city
Term 2005. 12 - 2007. 11
P.I Juntae Kim
본 과제는 협동적 추천 기술을 바탕으로 사용자와 선호 음악 사이의 관계를 분석하여 수백만 사용자, 수백만 아이템을 대상으로 실시간으로 음악 추천을 수행할 수 있는 그래프 모델을 기반으로 한 효율적인 음악 추천 엔진을 개발하고, 이를 국내 MP3 플레이어 선두업체인 아이리버의 ㈜ 레인콤과 협력하여 음악 포털 및 컨텐츠 유통에 적용, 사업화하는 것을 목표로 한다.
Adaptive Control of Multiple Characters in Client-Server Environment
(클라이언트-서버 환경에서 다중 캐릭터의 적응적 제어 기법 연구)
Funding Korea Science and Engineering Foundation
Term 2002. 9 - 2004. 8
P.I Kihyun Um
The goal of this research is to develop techniques for controlling the motion of multiple characters in the virtual space within a client/server environment. We will study methods of managing information about virtual space and the movement of characters, and develop the motion/path planing algorithm for artificial characters. The automatic generation of walking motions and the path panning algorithm can be used in a motion generator for 3D games. It can also be used in various Web-based application programs such as virtual shopping malls, interactive lecture systems, 3D games, etc.
Research on the customer preference prediction at the internet shopping mall (Completed)
(인터넷 쇼핑몰에서의 고객 기호 예측에 관한 연구)
Funding Korea Science and Engineering Foundation
Term 1999. 9 - 2001. 8
P.I Juntae Kim
As the electronic commerce market is rapidly growing, personalized product recommendation systems that suggest products for customers by analyzing each customer's personal preference are taking increasing interests. For the personalized recommendation, the ability of learning and predicting customer's preference is essential.
In this projectl, we study on the customer preference prediction method based on the customer behavior monitoring and collaborative filtering. We also propose the development of personalized recommendation system prototype to prove the usability of proposed method.
A development of medical image management system that provides diagnostic information by using medical knowledge base (Completed)
(의료 지식 베이스를 이용하여 진단 정보를 제공하는 의료 영상 관리 시스템 개발)
Funding Korea Science and Engineering Foundation
Term 1999. 9 - 2002. 8
P.I Kihyun Um
Recently, the picture archiving and communication systems (PACS) are widely used due to the growth of computer and communication technology. The PACS are more economic, easier to manage, and pollution-free compared to the film-based systems. Various commercial systems have been developed and used not only by foreign companies but also by domestic industry. However, the functionality of the current systems is limited to very basic operations such as storage and transmission of raw image data. In this research, we propose the development of the image processing technique to extract content-based features by using the knowledge base about human body structures, the way of constructing a medical image database with various diagnostic information in addition to the extracted features, and intelligent retrieval techniques with high precision.
A study on Web document ranking method based on hyperlink information and collaborative rating (Completed)
(하이퍼링크 정보와 협동적 평가를 기반으로한 웹 문서의 순위 결정 기법에 관한 연구)
Funding Ministry of Information and Communication
Term 1999. 7 - 2001. 6
P.I Juntae Kim
As the amount of information in Web rapidly grows, problems arise in current content-based search engines since the search result for simple keywords easily counts thousands with same similarity. In this research, we develop a search engine that reflects popularity and representativeness to the search result, by introducing a weighting method based on Collaborative rating and Hyperlink Information.
Development of Personalization Agent for Internet Portal Site (Completed)
(인터넷 포털 사이트를 위한 개인화된 에이전트의 개발)
Funding Ministry of Science and Technology, Netville Co.
Term 1999. 8. - 2000. 8.
P.I Juntae Kim
In this project, we have developed a software agent that learns user's interests in a portal site to dynamically provides personalized information. The machine learning techniques and IR techniques are used for Contend-based Filtering Agent and Collaborative Filtering Agent.
Personal Web information collection system based on compound profile learning (Completed)
(복합적인 프로파일 학습에 기반한 개인용 웹 정보 수집 시스템)
Funding Ministry of Information and Communication
Term 1997. 7 - 1999. 6
P.I Juntae Kim
The World Wide Web provides vast amount of information. As the size of the Web rows larger and larger, it becomes harder for ordinary people to find information that they need from the Web. Various kinds of software agents have been studied and proposed to assist people in searching for the Web. These agents include meta search engines, news and mail filtering systems, browsing assistants, and web page recommendation systems. In this project, a web agent called BookmarkFeeder has been developed that recommends web pages for a user. The BookmarkFeeder has two distinguished features compared with existing web agents. First, it uses implicit feedback for learning. It learns the users interests by monitoring the user behavior on the bookmark items without relying on explicit feedback. Second, it uses URL-based recommendation in conjunction with content-based recommendation for filtering web pages. By using the hyperlink information, a web page that contains no text can also be recommended.
Classification and intelligent indexing of text using knowledge-based natural language processing (Completed)
(지식기반 자연어처리를 이용한 문서 분류 및 지능형 색인에 관한 연구)
Funding Korea Science and Engineering Foundation
Term 1996. 3 - 1998. 2
P.I Juntae Kim
The importance of Information Retrieval and Information Extraction has been well recognized due to the rapid increase of online information sources and the growth of computer communications. Research on Information Retrieval has been focused on statistical models. Recently, however, there have been many attempts to apply natural language processing techniques because the statistical methods have limitations due to the lack of meaning analysis. On the other hands, in Information Extraction research, AI-based natural language processing techniques were broadly used, but current research direction is being changed to the use of statistical information, because of the difficulty in practical application of AI-based techniques. In this research, concerning these current trends, the development of text classification and indexing model, which increases the information retrieval performance by applying natural language processing technique based on pattern matching to the statistical information retrieval model, is proposed.
Development of Optimizing Class Scheduler (Completed)
(최적화된 강의 시간표 작성 시스템)
Funding Dongguk Univeristy
Term 1995. 10 - 1997. 9
P.I Juntae Kim
In this project, the automatic class scheduling program has been developed. The program is based on the Genetic Algorithm, in which each individual(timetable) is represented as two dimensional array(space-time). The program has been applied to scheduling the entire classes of Dongguk University Seoul Campus during the period of 1997 and 1998.