• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2005.11a
• /
• pp.110-113
• /
• 2005

본 논문은 가중 퍼지소속함수 기반 신경망(Neural Network with Weighted fuzzy Membership Funcstions, NEWFM)을 이용하여 심전도 신호로부터 조기심실수축(Premature Ventricular Contraction, PVC)을 판별하는 퍼지규칙을 추출하고 있다. NEWFM은 자기적응적(self adaptive) 가중 퍼지소속함수를 가지고 주어진 입력 데이터로부터 학습하여 퍼지규칙을 생성하고 이를 기반으로 정상 파형과 PVC 파형을 구분한다. 분류 성능 평가를 위하여 MIT/BIH 부정맥 데이터 베이스를 사용하였으며, NEWFM의 입력은 심전도의 파형에 웨이블릿 변환을 적용하여 추출된 웨이블릿 계수를 사용하였다. 여기에 비중복면적 분산 측정법을 적용하여 중요도가 낮은 계수를 제거하면서 최소의 m 개 특징입력만을 사용한 하이퍼박스로 단순화 시킨다. 이러한 방법으로 추출된 2개의 웨이블릿 계수를 사용한 퍼지규칙은 $96\%$의 PVC 분류성능을 보여준다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.5
• /
• pp.482-488
• /
• 2014

This paper aims to analyze user's emotion automatically by analyzing Twitter, a representative social network service (SNS). In order to create sentiment analysis models by using machine learning techniques, sentiment labels that represent positive/negative emotions are required. However it is very expensive to obtain sentiment labels of tweets. So, in this paper, we propose a sentiment analysis model by using self-training technique in order to utilize "data without sentiment labels" as well as "data with sentiment labels". Self-training technique is that labels of "data without sentiment labels" is determined by utilizing "data with sentiment labels", and then updates models using together with "data with sentiment labels" and newly labeled data. This technique improves the sentiment analysis performance gradually. However, it has a problem that misclassifications of unlabeled data in an early stage affect the model updating through the whole learning process because labels of unlabeled data never changes once those are determined. Thus, labels of "data without sentiment labels" needs to be carefully determined. In this paper, in order to get high performance using self-training technique, we propose 3 policies for updating "data with sentiment labels" and conduct a comparative analysis. The first policy is to select data of which confidence is higher than a given threshold among newly labeled data. The second policy is to choose the same number of the positive and negative data in the newly labeled data in order to avoid the imbalanced class learning problem. The third policy is to choose newly labeled data less than a given maximum number in order to avoid the updates of large amount of data at a time for gradual model updates. Experiments are conducted using Stanford data set and the data set is classified into positive and negative. As a result, the learned model has a high performance than the learned models by using "data with sentiment labels" only and the self-training with a regular model update policy.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.7 no.4
• /
• pp.905-915
• /
• 2012

This paper suggests the online method to identify normal and abnormal state of water quality on the ocean USN. To define normal of the ocean water quality, we utilize the negative selection algorithm of artificial immunity system which has self and nonself identification characteristics. To distinguish abnormal status, normal state set of the ocean water quality needs to be defined. For this purpose, we generate normal state set base on mutations of each data and mutation of the data as logical product. This mutated normal (or self) sets used to identify abnormal status of the water quality. We represent the experimental result about mutated self set with the Gaussian function. Through setting the method on the ocean sensor logger, we can monitor whether the ocean water quality is normal or abnormal state by online.

• Current Industrial and Technological Trends in Aerospace
• /
• v.4 no.1
• /
• pp.83-91
• /
• 2006

추적레이다는 표적으로부터 반사되어 돌아오는 신호 또는 질의 신호에 대한 응답 신호를 수신하여 표적을 추적하는 장비이다. 추적레이다가 표적을 추적하는 범위는 일반적으로 좁게 한정되므로 이동하는 표적을 추적하기 위해서는 먼저 안테나 빔의 지향각과 거리를 표적에 맞추고, 표적이 획득된 후에는 안테나 빔을 연속적으로 이동하는 표적을 향해 방사하여 표적을 추적하게 된다. 일반적으로 추적레이다가 표적을 추적하는 경우에는 과정 잡음과 측정 잡음에 의해서 발생되는 부정확성과 관심없는 표적이나 클러터 등으로부터 생성된 측정 근원의 부정확성으로 인한 문제가 발생하게 된다. 이러한 표적 추적에 따른 문제를 해결하기 위해서 많은 추적 알고리듬들이 개발되어 왔다. 이 논문에서는 가장 기본적인 표준 칼만 필터와 측정 근원의 부정확성에 따른 데이터 연관 문제를 고려한 여러 추적 알고리듬에 대해서 기술하였다. 또한 한국항공우주연구원 우주센터의 우주발사체 추적용 추적레이다에 대한 간략한 설명과 우주발사체 추적에 사용되는 추적 알고리듬에 대해서 소개하였다.

• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.400-403
• /
• 2019

스타일 변환(Style Transfer)은 주어진 문장의 긍정이나 부정 같은 속성을 변경하여 다른 속성을 갖는 문장으로 변환하는 과정을 의미한다. 본 연구에서는 스타일 변환을 위한 단어 n-그램 삭제의 기준을 확장하였고, 네이버 영화리뷰 데이터셋을 통해 이를 스타일 변환 이후 원래 문장의 스타일로부터 얼마나 차이가 나게 되었는지를 측정하였다. 측정은 감성분석기를 통해 이루어졌고, 기존 방법에 비해 6.28%p정도 높은 75.13%의 정확도를 보였다.

• Journal of Internet of Things and Convergence
• /
• v.9 no.1
• /
• pp.77-87
• /
• 2023

Applying various association rule mining algorithms to the network intrusion detection task involves two critical issues: too large size of generated rule set which is hard to be utilized for IoT systems and hardness of control of false negative/positive rates. In this research, we propose an association rule mining algorithm based on the newly defined measures called coverage and exclusion. Coverage shows how frequently a pattern is discovered among the transactions of a class and exclusion does how frequently a pattern is not discovered in the transactions of the other classes. We compare our algorithm experimentally with the Apriori algorithm which is the most famous algorithm using the public dataset called KDDcup99. Compared to Apriori, the proposed algorithm reduces the resulting rule set size by up to 93.2 percent while keeping accuracy completely. The proposed algorithm also controls perfectly the false negative/positive rates of the generated rules by parameters. Therefore, network analysts can effectively apply the proposed association rule mining to the network intrusion detection task by solving two issues.

• Journal of Korea Society of Industrial Information Systems
• /
• v.28 no.5
• /
• pp.15-30
• /
• 2023

Recently, generative models based on the Transformer architecture, such as ChatGPT, have been gaining significant attention. The Transformer architecture has been applied to various neural network models, including Google's BERT(Bidirectional Encoder Representations from Transformers) sentence generation model. In this paper, a method is proposed to create a text binary classification model for determining whether a comment on Korean movie review is positive or negative. To accomplish this, a pre-trained multilingual BERT sentence generation model is fine-tuned and transfer learned using a new Korean training dataset. To achieve this, a pre-trained BERT-Base model for multilingual sentence generation with 104 languages, 12 layers, 768 hidden, 12 attention heads, and 110M parameters is used. To change the pre-trained BERT-Base model into a text classification model, the input and output layers were fine-tuned, resulting in the creation of a new model with 178 million parameters. Using the fine-tuned model, with a maximum word count of 128, a batch size of 16, and 5 epochs, transfer learning is conducted with 10,000 training data and 5,000 testing data. A text sentiment binary classification model for Korean movie review with an accuracy of 0.9582, a loss of 0.1177, and an F1 score of 0.81 has been created. As a result of performing transfer learning with a dataset five times larger, a model with an accuracy of 0.9562, a loss of 0.1202, and an F1 score of 0.86 has been generated.

• Proceedings of the Korean Society of Propulsion Engineers Conference
• /
• 2005.11a
• /
• pp.299-303
• /
• 2005

In the present study, in order to improve precision of the component characteristic maps generated by the scaling method, a map generation method which can produce a compressor map from some experimental performance data using GAs(Genetic Algorithms) was proposed. However, in case of the proposed map generation method only using GAs, because it has a drawback for estimating correctly the surge points and the choke points of the compressor map, a modified GAs method was additionally proposed through complementally use of the scaling method to determine obviously those points of the compressor map.

• Journal of Korea Multimedia Society
• /
• v.16 no.1
• /
• pp.67-74
• /
• 2013

In this paper, we propose the pointing and correction algorithm for optimized performance based on Bluetooth communication. The error from the accelerometer sensor's output must be carefully managed as the accelerometer sensor is more sensitive to data change compared to that of the gyroscope sensor. Thus, we minimize the noise by applying the Kalman filter to data for each axis from the accelerometer. In addition, we can also obtain effect compensating the hand tremor by applying the Kalman filter to the data variation for x and y. In this study, we extract data through the Quaternion mapping process on data from the accelerometer and gyroscope. In turn, we can obtain a tilt compensation by applying a compensation algorithm with acceleration of the gravity of the extracted data. Moreover, in order to correct the inaccuracy on smart sensor due to the rapid movement of a device, we propose a adaptive pointing and correction algorithm using the genetic approach to generate the initial population depending on the user.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.10a
• /
• pp.1030-1032
• /
• 2019

형태소 분석을 통하여 한국어 문장을 형태소 단위의 임베딩 및 학습 관련 연구가 되었으나 최근 비정형적인 텍스트 데이터의 증가에 따라 음소 단위의 임베딩을 통한 신경망 학습에 대한 요구가 높아지고 있다. 본 논문은 비정형적인 텍스트 감성 분석 성능 향상을 위해 음소 단위의 토큰을 생성하고 이를 CNN 모형을 기반으로 다차원 임베딩을 수행하고 감성분석을 위하여 양방향 순환신경망 모델을 사용하여 유튜브의 비정형 텍스트를 학습시켰다. 그 결과 텍스트의 긍정 부정 판별에 있어 90%의 정확도를 보였다.