• KIPS Transactions on Computer and Communication Systems
• /
• v.10 no.9
• /
• pp.251-260
• /
• 2021

Gait analysis is one of the research fields for obtaining various information related to gait by analyzing human ambulation. It has been studied for a long time not only in the medical field but also in various academic areas such as mechanical engineering, electronic engineering, and computer engineering. Efforts have been made to determine whether there is a problem with gait through gait analysis. In this paper, as a pre-step to find out gait abnormalities, it is investigated whether it is possible to differentiate whether experiment participants wear elderly simulation suit or not by applying gait data to machine learning models for the same person. For a total of 45 participants, each gait data was collected before and after wearing the simulation suit, and a total of six machine learning models were used to learn the collected data. As a result of using an artificial neural network model to distinguish whether or not the participants wear the suit, it showed 99% accuracy. What this study suggests is that we explored the possibility of judging the presence or absence of abnormality in gait by using machine learning.

• Journal of Korea Multimedia Society
• /
• v.17 no.2
• /
• pp.232-239
• /
• 2014

As online texts have been rapidly growing, their automatic classification gains more interest with machine learning methods. Nevertheless, comparatively few research could be found, aiming for Korean texts. Evaluating them with statistical methods are also rare. This study took a sample of tweets and used machine learning methods to classify emotions with features of morphemes and n-grams. As a result, about 76% of emotions contained in tweets was correctly classified. Of the two methods compared in this study, Support Vector Machines were found more accurate than Na$\ddot{i}$ve Bayes. The linear model of SVM was not inferior to the non-linear one. Morphological features did not contribute to accuracy more than did the n-grams.

• The Journal of Society for e-Business Studies
• /
• v.25 no.3
• /
• pp.1-13
• /
• 2020

Machine learning refers to a model generation technique that can solve specific problems from the generalization process for given data. In order to generate a high performance model, high quality training data and learning algorithms for generalization process should be prepared. As one way of improving the performance of model to be learned, the Ensemble technique generates multiple models rather than a single model, which includes bagging, boosting, and stacking learning techniques. This paper proposes a new Ensemble technique with multiple stacking that outperforms the conventional stacking technique. The learning structure of multiple stacking ensemble technique is similar to the structure of deep learning, in which each layer is composed of a combination of stacking models, and the number of layers get increased so as to minimize the misclassification rate of each layer. Through experiments using four types of datasets, we have showed that the proposed method outperforms the exiting ones.

• Journal of the Korean earth science society
• /
• v.45 no.2
• /
• pp.111-120
• /
• 2024

Cloud coverage is a key factor in determining whether to proceed with observations. In the past, human judgment played an important role in weather evaluation for observations. However, the development of remote and robotic observation has diminished the role of human judgment. Moreover, it is not easy to evaluate weather conditions automatically because of the diverse cloud shapes and their rapid movement. In this paper, we present the development of a cloud monitoring program by applying a machine learning-based Python module "cloudynight" on all-sky camera images obtained at Miryang Arirang Astronomical Observatory (MAAO). The machine learning model was built by training 39,996 subregions divided from 1,212 images with altitude/azimuth angles and extracting 16 feature spaces. For our training model, the F1-score from the validation samples was 0.97, indicating good performance in identifying clouds in the all-sky image. As a result, this program calculates "Cloudiness" as the ratio of the number of total subregions to the number of subregions predicted to be covered by clouds. In the robotic observation, we set a policy that allows the telescope system to halt the observation when the "Cloudiness" exceeds 0.6 during the last 30 minutes. Following this policy, we found that there were no improper halts in the telescope system due to incorrect program decisions. We expect that robotic observation with the 0.7 m telescope at MAAO can be successfully operated using the cloud monitoring program.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1994.04a
• /
• pp.198-207
• /
• 1994

공작기계부품 가공을 위한 공정표는 가공공정, 공정별 도면 분할, 가공기계 등을 결정하는 공정계획과 한 공정에 대하여 가공방법, 가공공구, 절삭조건, 공수등을 결정하는 작업계획을 통하여 발행된다. 작업계획에서 가공방법과 가공공구의 결정은 절삭조건과 공수에 영향을 주는 중요한 요소이다. 기존의 연구에서는 가공방법과 가공공구를 결정하기 위해 전문가 시스템 쉘(expert system shell)이용한 사례가 많았다. 이 경우, 지식 베이스(knowledge base) 의 구축에 많은 시간이 소요되고, 지식이 변했을 때 수정의 어려움이 있다. 본 연구에서는 표준화되지 않아 변경의 소지가 많은 가공방법과 가공공구 결정에 뉴럴 네트워크(neural network)의 한 종류인 백 프로퍼게이션 (back propagation) 학습 모델을 이용했다. 공정계획 후 분할된 공정별 도면으로부 터 크기 및 정밀도 등과 같은 특징형상(feature) 정보를 추출한 후, 특징형상 의 종류와 크기, 치수공차, 기하공차, 거칠기 등을 입력하여 가공방법 및 가 공공구가 출력되도록 학습패턴을 설정하여 학습시켰다. 학습패턴은 공정설계 전문가와 인터뷰하는 방법과 작업계획 과정을 분석하는 방법을 통하여 설정 했다. 백 프로퍼게이션 모델을 통하여 학습시킨 결과, 학습시킨대로 정확한 가공방법 및 가공공구를 결정할 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2005.05a
• /
• pp.1119-1122
• /
• 2005

기계학습 방법을 이용한 네트워크 기반 침입탐지 시스템은 어떤 학습알고리즘을 사용하여 구현되었느냐에 따라 그 결과가 매우 달라진다. 학습을 위한 전처리를 많이 하면 비례하여 성능이 개선되지만, 실제 사용의 유용성면에서는 성능이 떨어지게 된다. 따라서 최소한의 전처리를 하여 침입탐지의 탐지율을 보장하는 방법이 필요 하다. 본 논문에서는 네트워크기반 침입탐지 문제를 기계학습을 이용하여 해결하는 방법을 제안 하였다. 제안된 모델은 탐지 속도와 각종 공격들의 패킷 분포를 고려하여 관련된 그룹으로 분류하고, 이것을 학습하는 시스템이다. 실험을 통하여 제안된 모델의 유용성을 검증 하였다.

• Journal of Convergence for Information Technology
• /
• v.11 no.6
• /
• pp.270-281
• /
• 2021

With the development of information technology, the size of insurance fraud is increasing rapidly every year, and the method is being organized and advanced in conspiracy. Although various forms of prediction models are being studied to predict and detect this, insurance-related information is highly sensitive, which poses a high risk of sharing and access and has many legal or technical constraints. In this paper, we propose a machine learning insurance fraud prediction model based on blockchain, one of the most popular technologies with the recent advent of the Fourth Industrial Revolution. We utilize blockchain technology to realize a safe and trusted insurance information sharing system, apply the theory of social relationship analysis for more efficient and accurate fraud prediction, and propose machine learning fraud prediction patterns in four stages. Claims with high probability of fraud have the effect of being detected at a higher prediction rate at an earlier stage, and claims with low probability are applied differentially for post-reference management. The core mechanism of the proposed model has been verified by constructing an Ethereum local network, requiring more sophisticated performance evaluations in the future.

• Journal of Digital Convergence
• /
• v.19 no.7
• /
• pp.199-208
• /
• 2021

Automatic Post Editing(APE) is the study that automatically correcting errors included in the machine translated sentences. The goal of APE task is to generate error correcting models that improve translation quality, regardless of the translation system. For training these models, source sentence, machine translation, and post edit, which is manually edited by human translator, are utilized. Especially in the recent APE research, multilingual pretrained language models are being adopted, prior to the training by APE data. This study deals with multilingual pretrained language models adopted to the latest APE researches, and the specific application method for each APE study. Furthermore, based on the current research trend, we propose future research directions utilizing translation model or mBART model.

• Phonetics and Speech Sciences
• /
• v.14 no.4
• /
• pp.35-43
• /
• 2022

This paper studies how to improve the identification rate of laryngeal disability speech data by convolutional neural network (CNN) and machine learning ensemble learning methods. In general, the number of laryngeal dysfunction speech data is small, so even if identifiers are constructed by statistical methods, the phenomenon caused by overfitting depending on the training method can lead to a decrease the identification rate when exposed to external data. In this work, we try to combine results derived from CNN models and machine learning models with various accuracy in a multi-voting manner to ensure improved classification efficiency compared to the original trained models. The Pusan National University Hospital (PNUH) dataset was used to train and validate algorithms. The dataset contains normal voice and voice data of benign and malignant tumors. In the experiment, an attempt was made to distinguish between normal and benign tumors and malignant tumors. As a result of the experiment, the random forest method was found to be the best ensemble method and showed an identification rate of 85%.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.269-274
• /
• 2023

질의응답 (Question Answering)은 주어진 질문을 이해하여 그에 맞는 답변을 생성하는 자연어 처리 분야의 핵심적인 기계 독해 작업이다. 현재 대다수의 자연어 이해 작업은 사전학습 언어 모델에 미세 조정 (finetuning)하는 방식으로 학습되고, 질의응답 역시 이러한 방법으로 진행된다. 하지만 미세 조정을 통한 전이학습은 사전학습 모델의 크기가 커질수록 전이학습이 잘 이루어지지 않는다는 단점이 있다. 게다가 많은 양의 파라미터를 갱신한 후 새로운 가중치들을 저장하여야 한다는 용량의 부담이 존재한다. 본 연구는 최근 대두되는 deep prompt tuning 방법론을 한국어 추출형 질의응답에 적용하여, 미세 조정에 비해 학습시간을 단축시키고 적은 양의 파라미터를 활용하여 성능을 개선했다. 또한 한국어 추출형 질의응답에 최적의 prompt 길이를 최적화하였으며 오류 분석을 통한 정성적인 평가로 deep prompt tuning이 모델 예측에 미치는 영향을 조사하였다.