스팸메일 포착을 위한 데이터마이닝 기법 비교

김동건,조진남,장도석 同德女子大學校 情報科學硏究所 2004 정보과학연구 Vol.8 No.-

스팸메일로 인한 사회적 손실이 급증하고 있는 가운데 스팸메일을 포착하는 각종 방법들이 개발되고 있다. 본 연구에서는 스팸메일엔진에 사용 가능한 데이터마이닝기법을 단일기법과 혼합기법으로 나누어 Spambase데이터에 적용하였다. 적용 결과, 가장 많이 쓰이고 있는 Naive Bayes기법의 성능이 가장 저조하였던 반면, SVM과 복합기법의 성능이 탁월한 것으로 나타났다. This study compares recently developed data mining algorithms for detecting spam mails. We have considered single classifiers, such as Naive Bayes classifier, LDA, QDA, logistic discriminant, and support vector machines and ensemble methods such as Bagging and Random Forest. While SVM and ensemble methods show excellent performance, Naive Bayes classifier which has been popular in detecting spam mails shows low performance.

사용자 인식 스마트 홈 미러

최건영,고성화,김종옥,박은서,김완규,백경동,박동규 순천향대학교 부설 산업기술연구소 2021 순천향 산업기술연구소논문집 Vol.27 No.2

The Internet of Things, the core of the Fourth Industrial Revolution, is growing by incorporating various technologies into various fields. Among them, smart home systems that apply IoT technology to home environments are being actively studied. The home hub market, a means of communication for smart homes, is also developing at the same time. As the home hub market develops, security threats are also increasing accordingly. In this paper, 'Smart Home Mirror' is implemented by combining mirror and IoT technologies so that users can use them to communicate with smart home applications and other devices. To meet security issues, users in it are recognized by using face recognition technology and AES between all data communications in it is applied. The proposed device’s usefulness is approved by implementing it in smart home environments.

최소제곱혼합분포추정량을 이용한 판별분석

김동건,조진남 同德女子大學校 情報科學硏究所 2005 정보과학연구 Vol.9 No.-

최소제곱혼합분포추정(The Least Squares Mixture Decomposition Density Estimator)을 K개의 부모집단 분포추정에 적용하여 판별분석으로 확장하였다. 기존의 커널판별분석과 비교한 결과 모형은 커널판별분석보다 현저히 단순해진 반면, 예측력은 커널판별분석과 거의 동일한 수준이었다. The least squares mixture decomposition density estimation has been extended and applied to the discrimant analysis for K subpopulation classification. The complexity of the classification model by the least squares mixture decomposition is considerably reduced compared to that of the classification by kernel method. The performance of the least squares mixture decomposition method is comparable with the kernel method.

水 海洋係 專門大學 通信科 敎育課程 開發 硏究

金永煜,金秀植,高南泳,梁義藏,金圭煥,文建,鄭東洙 군산대학교 1984 群山水産專門大學 硏究報告 Vol.18 No.1

We have studied on the curriculum to develop the communication department of the inter-related fisheries and marine junior college groups and abridged them as follows. (1) A vessel communication has the special characteristics compared with general communication field and also it has an aim to protect the lives and the property from dangerous navigation. For these reasons we must educate many competent graduates who are versed in business and accustom themselves to necessary knowledge of communication and can be adapted them selves to the advanced communication method, techniques and international information on the communication. (2) Training goal and way of thinking of students were rearranged and the occupational cluster in the electronic-communication field was fractionated to develop the curriculum. Especially regarding operator's duty served the above purpose as a occupational cluster, we analyzed their job in accordance with it. (3) International convention on standards of training certification and watchkeeping for seafarers, 1978 was reflected, also international telecommunication convention was referred so that it might be satisfied with the condition to be observed. (4) Developing the curriculum, we responed and analyzed the questions from scientific world, industrial organizations and graduates and provide their advice and suggestions form consultive committee and reflected them to develop the curriculum

정신분열병 또는 정신분열형장애를 가진 입원환자에서 Risperidone과 Olanzapine의 효과의 비교 : 6주간의 개방적 전향적 연구 6 Weeks, Open-Labelled, Prospective Study

김찬우,이동건,강승훈,조지현,이수진,조동환,김정기 대한생물치료정신의학회 2003 생물치료정신의학 Vol.9 No.2

Objectives : Risperidone and olanzapine, which are atypical antipsychotics. are widely used in the treatment of schizophrenia as the 1st line drug, proving their effectiveness for both positive and negative symptoms. They also have many advantages even from the side effects. However, they show differences with the chemical structure and pharmacokinetics and they work differently to receptors. Researchers have studied the effect and side effect during the acute phase of schizophrenia as an open prospective study. Methods : Forty subjects with DSM-IV schizophrenia or schizoaffective patients including 1) 1st onset schizophrenia patient, 2) patient who has been treated by the prior antipsychotics and not improved or even worse to be admitted again, were assigned to receive risperidone or olanzapine for 6 weeks. Psychopathology, adverse effect. weight change, fasting serum glucose, and drug cost were compared within and across groups. Results : After 6 weeks later both treatment groups showed the significant decrease in PANSS positive, general psychopathology and total score while they did not show any significant difference between two groups. Risperidone group, however. showed the significant decrease in PANS negative score, which olanzapine group did not. In terms of adverse effect risperidone group showed more extrapyramidal symptom than olanzapine group. Body weight rised significantly after Gweeks, but no significant difference between two treatment groups. There were no significant indrease of fasting serum glucose level. Conclusions : Risperidone and olanzapine are effective drugs which can be used during the acute phase of schizophrenia. Risperidone is expected to shorten the admission day and remission of the symptom of acute phase, mile olarizapine is more advantage in a neurologic side effect than rispendone.

실크 생지직물의 정련발염날염에 관한 연구

박건용,박창혁,서기성,김재현,김동철 호서대학교 반도체제조장비국산화연구센터 2001 반도체장비학술심포지움 Vol.2001 No.-

실크 생지직물에 대해 정련날염호로 무늬를 인날하고 증열 처리함으로써 무늬 부분만 정련이 되도록 하는 정련날염 기술을 개발하여 모시 섬유와 같이 뻣뻣하면서 시원한 질감이 나는 정련되지 않은 생직물 부분과 정련에 의해 유연하고 매끄러우면서 광택이 나는 비단 부분이 동시에 한 직물에 공존하면서 입체적인 무늬 표현이 가능한 독특하고 새로운 실크 직물을 개발하였다. 또한 산성염로 등의 합성염료로 염색된 실크 생지직물과 천연염fy로 염색된 실크 생지직물에 대해서 정련날염과 동시에 무늬 부분의 바탕색을 빼내고 착색을 고착시키는 발염의 특수날염 기술을 개발하여 다양한 색상에 의한 무늬 표현이 가능하게 함으로써 고부가가치 실크직물을 생산할 수 있도록 되었다.

AHP와 QFD를 이용한 계층별 성과평가의 측정지표 개발과 이익효과에 관한 연구

兪日根,金東垣 弘益大學校 科學技術硏究所 1998 科學技術硏究論文集 Vol.9 No.2

Today, companies try to adopt new management technologies to survive in changing business environment. But it results in a few successful cases. The reason for it is that they do not realize the exact problems of the company and they apply just the techniques without proper modification, and also they just analyze only financial factor for performance evaluation of their corporate by use of only the consequence indexes but the process measurement indexes. This paper analyzes these problem and proposes the method of improvement. That is, this paper proposes new efficient methodology and model which can evaluate not only the factor of management but also the process of the management. A company organization can be divided for evaluation index into three, those are, organization level, process level, and team/individual level. AHP and QFD methods are applied to find the critical performance measurement indexes which meet the basic selecting conditions. The critical measurement indexes would work as essential factors of evaluation system in the priority order chosen by the goal of a company. This paper proposes 5 phases of new evaluation system model. And the process of team evaluation indexes are developed and case studies are represented based on the above method. When the evauation indexes are selected, corporate organization can be divided into three levels: 1) organization level 2) process level 3) team/personal level. The index type is composed with three categories: 1) cost 2) quality 3) time. It must satisfy the customer, employee and quality systems requirement. In conclusion, the five phases for the new evaluation system are proposed in this study and also the processes to develop the system evaluating model especially for the team level and sample case studies are represented.

가공열처리에 의한 Ti-6Al-4V 합금의 조직 및 경도 변화에 대한 연구

서동이,김성진,한전건 成均館大學校 科學技術硏究所 1988 論文集 Vol.39 No.2

The microstructure and hardness change by thermomechanical processing was studied in the temperature range of α+β phase(800℃) and above the β-transus temperature(1,050℃) respectively in Ti-6A1-4V alloy. Equiaxed a morphology could be obtained through hot rolling and annealing at 800℃ while a lamellar structure was produced only by solution treatment followed by annealing without hot rolling. The morphology change from lamellar a to equiaxed a was Promoted with increasing the degree of deformation. Thermome chanical processing above β transus temperature resulted in the combined structure of serrated α, acicular a and fine lamellar structure. It was also found that prolonged annealing could induce the formation of egaiaxed α at highly stressed area. Hardness appeared to be higher in lamellar a structure than equiaxed a structure.

변량 계수 모형의 통계적 성질과 실제 적용 모형에 관한 연구

조진남,김동건 同德女子大學校 情報科學硏究所 2005 정보과학연구 Vol.9 No.-

생물을 대상으로 의학, 약학 등의 실험에서 개체별 처리 시점이 일정하지 않을 경우, 시간과 반응값의 관계를 규명하고자 할 때 변량계수모형을 적용한다. 이러한 관계는 선형 또는 2차, 3차 함수의 다항식의 형태로 표현된다. 변량계수 모형은 혼합모형에 속하므로 변량계수모형의 추정과 검정은 혼합모형의 추정 및 검정절차에 따른다. 그리고 실제 실험에서 흔히 적용되는 변량계수모형들의 모형식과 통계적 성질을 규명한다. A random coefficient model is applied when Times of the repeated measures of data are not fixed in experiments of medicine, pharmaceutics with respect to subjects. The model is devised to describe arithmetically the relationship of a measurement with time, especially in linear, quadratic, or cubic relationship. The random coefficient model is considered to be one of mixed models. Hence, the procedures of the inference of a random coefficient model are same as those of a mixed model. Finally, we investigate the equations of various models and statistical properties followed by the models in random coefficient models.

Cheeking Estimability in General Linear Models using Mathematica

조진남,김동건 同德女子大學校 1997 同大論叢 Vol.27 No.1

통계학에서의 중요한 결과들은 많은 대수적인 연사을 거쳐서 얻어진다. 통계학을 전공하거나 이용하는 사람들은 많은 수식 조작을 하는 과정에서 지엽적인 수식 조작에 얽매여 결론에 도달하기도 전에 포기하거나 결론 자체를 이해하지 못하는 경우가 많다. 이런 수식 조작 과정은 Mathematica 등의 수식 조작이 가능한 프로그램(symbolic manipulation software)을 통해서 상당 부분 생략될 수 있다. 여기서는 Mathematica를 이용하여 선형 연립방정식이 기초적인 결과로부터 일반 성형 모형에서의 추정 가능 함수를 찾거나 또는 어떤 특정 모수의 선형 결합이 추정 가능한지를 검사하는 방법을 제시하고, 일원 분산 분석과 이원 분산 분석을 실례로 들어 설명한다.