Earticle

본 논문에서는 기존에 전문가에 의해서 이루어지던 국가 대기오염 측정망 데이터들의 이상 탐지 작업을 인공지능을 통해 자동화하고자 심층 신경망을 이용한 이상 탐지 모델을 제안하였다. 환경과학원에서 제공받은 기상자료 데이터의 결측치 및 이상치를 분석하여 학습데이터를 생성하였으며 비지도 학습 방식의 BeatGAN 모델에 기반하여 커널 구조 변경과 합성곱 필터층 및 전치 합성곱 필터층의 추가를 통해 새로운 모델을 제안하여 이상 탐지 성능을 높이고자 하였다. 또한 제안하는 모델의 생성적 특징을 활용하여 새로운 데이터를 생성하고 이를 학습에 사용하는 재학습 알고리즘을 구현 및 적용하여 기존 BeatGAN 모델뿐 아니라 다른 비지도 학습 모델인 Iforest, One Class SVM과 비교하였을 때 제안모델의 성능이 가장 높았음을 확인할 수 있었다. 본 연구를 통해 실제 산업현장에서 센서의 이상, 점검 등의 여러 요인으로 인해 학습 데이터가 부족한 상황에서 추가 적인 비용없이 과적합을 피하며 제안하는 모델의 이상탐지 성능을 올릴 수 있는 방법을 제시할 수 있었다.

In this paper, We propose an anomaly detection model using deep neural network to automate the identification of outliers of the national air pollution measurement network data that is previously performed by experts. We generated training data by analyzing missing values and outliers of weather data provided by the Institute of Environmental Research and based on the BeatGAN model of the unsupervised learning method, we propose a new model by changing the kernel structure, adding the convolutional filter layer and the transposed convolutional filter layer to improve anomaly detection performance. In addition, by utilizing the generative features of the proposed model to implement and apply a retraining algorithm that generates new data and uses it for training, it was confirmed that the proposed model had the highest performance compared to the original BeatGAN models and other unsupervised learning model like Iforest and One Class SVM. Through this study, it was possible to suggest a method to improve the anomaly detection performance of proposed model while avoiding overfitting without additional cost in situations where training data are insufficient due to various factors such as sensor abnormalities and inspections in actual industrial sites.

뇌파 파형중 베타파를 이용한 인간의 인지상태를 판별한다. 베타파는 인간의 인지상태중 스트레스 영역에 해당하는 특성이 있고, 이 영역에서 스트레스의 오버대역폭을 추출하기 위해서 저대역폭과 고대역폭 사이의 베타파간 상관관계를 분 석해야 한다. 그러므로 본 논문에서는 효과적으로 베타파 상관관계를 분석하고 추출하기 위해 비지도학습 머신러닝을 이용 한 Kmean 클러스터링 분석모델을 제시한다. 제시된 모델은 베타파 영역을 유사한 영역의 클러스터 군으로 분류하고 해당 클러스터링 범주에서 이상파형을 판별한다. 이상파형 판별을 위해 클러스터군의 밀집도와 정상범주 이탈영역을 기준으로 스트레스 위험군을 판별하고 판별된 스트레스 위험군에 대한 대처방안을 제공할 수 있다. 제시된 모델을 활용하면 뇌파파형 을 통한 인지상태의 스트레스 지수분별이 가능하고, 개인의 인지상태에 대한 관리 및 응용이 가능하다. 또한 스트레스와 오피스증후군을 갖는 사람들에게 뇌파관리를 통해 개인의 삶에 대한 질적 향상에 도움을 준다.

The characteristic of the beta wave among the EEG waves corresponds to the stress area of human perception. The over-bandwidth of the stress is extracted by analyzing the beta-wave correlation between the low-bandwidth and high-bandwidth. We present a KMeans clustering analysis model for unsupervised machine learning to construct an analytical model for analyzing and extracting the beta-wave correlation. The proposed model classifies the beta wave region into clusters of similar regions and identifies anomalous waveforms in the corresponding clustering category. The abnormal group of waveform clusters and the normal category leaving region are discriminated from the stress risk group. Using this model, it is possible to discriminate the degree of stress of the cognitive state through the EEG waveform, and it is possible to manage and apply the cognitive state of the individual.

We present an unsupervised anomaly detection framework for aluminum molten metal processes that trains exclusively on normal operational data. The extreme temperatures and dynamic luminance variations characteristic of molten metal environments make it practically infeasible to collect representative samples of anomalies—such as foreign material contamination, thermal irregularities, or flow inconsistencies—and manual annotation remains prohibitively labor-intensive. Our approach employs a Convolutional Autoencoder to capture the visual and thermal signatures of normal process states. Anomalies are detected by computing reconstruction errors and comparing them against thresholds derived from the upper percentiles of the normal error distribution. Experimental validation across multiple production lines demonstrates that our method achieves high detection accuracy even with limited abnormal samples, offering a practical solution for automated quality control and real-time process monitoring in industrial casting operations.

본 논문에서는 사용자의 EEG(Electroencephalogram)신호를 측정하여 SVM(Support Vector Machine) 알 고리즘을 이용하여 EEG 신호룰 분류하고 신호의 정확도를 측정하였다. 사용자의 EEG 신호를 측정하기 위해 남·여를 구분하여 실험을 진행하였으며, EEG 신호 측정은 단채널 EEG 디바이스를 이용하였다. EEG 디바이스를 이용하여 사용 자의 EEG 신호를 측정한 결과는 R을 이용하여 분석하였다. 또한 SVM의 분류 성능이 최고가 되는 특정 벡터의 조합을 적용시켜 EEG 측정 실험 데이터를 80:20(훈련 데이터: 테스트 데이터) 비율로 예측해 본 결과 인식률 93.2% 의 예측 정확도를 보였다. 본 논문에서는 사용자의 EEG 신호를 약 93.2% 정도로 인식할 수 있었으며, SVM 알고리즘의 간단한 선형 분류만으로 수행이 가능하다는 점은 EEG 신호를 이용하여 생체인증에 다양하게 활용될 수 있음을 제시하였다.

In this paper, we measured the user's EEG signal and classified the EEG signal using the Support Vector Machine algorithm and measured the accuracy of the signal. An experiment was conducted to measure the user's EEG signals by separating men and women, and a single channel EEG device was used for EEG signal measurements. The results of measuring users' EEG signals using EEG devices were analyzed using R. In addition, data in the study was predicted using a 80:20 ratio between training data and test data by applying a combination of specific vectors with the highest classifying performance of the SVM, and thus the predicted accuracy of 93.2% of the recognition rate. This paper suggested that the user's EEG signal could be recognized at about 93.2 percent, and that it can be performed only by simple linear classification of the SVM algorithm, which can be used variously for biometrics using EEG signals.

비강체 정합은 임상적 필요성은 높으나 계산 복잡도가 높고, 정합의 정확성 및 강건성을 확보하기 어려운 분야이다. 본 논문은 비지도 학습 환경에서 3차원 뇌 자기공명 영상 데이터에 딥러닝 네트워크를 이용한 비강체 정합 기법을 제 안한다. 서로 다른 환자의 두 영상을 입력받아 네트워크를 통하여 두 영상 간의 특징 벡터를 생성하고, 변위 벡터장을 만들어 기준 영상에 맞추어 다른 쪽 영상을 변형시킨다. 네트워크는 U-Net 형태를 기반으로 설계하여 정합 시 두 영상의 전역적, 지역적인 차이를 모두 고려한 특징 벡터를 만들 수 있고, 손실함수에 균일화 항을 추가하여 3차원 선 형보간법 적용 후에 실제 뇌의 움직임과 유사한 변형 결과를 얻을 수 있다. 본 방법은 비지도 학습을 통해 임의의 두 영상만을 입력으로 받아 단일 패스 변형으로 비강체 정합을 수행한다. 이는 반복적인 최적화 과정을 거치는 비학습 기반의 정합 방법들보다 빠르게 수행할 수 있다. 실험은 50명의 뇌를 촬영한 3차원 자기공명 영상을 가지고 수행하 였고, 정합 전·후의 Dice Similarity Coefficient 측정 결과 평균 0.690으로 정합 전과 비교하여 약 16% 정도의 유사도 향상을 확인하였다. 또한, 비학습 기반 방법과 비교하여 유사한 성능을 보여주면서 약 10,000배 정도의 속도 향상을 보여주었다. 제안 기법은 다양한 종류의 의료 영상 데이터의 비강체 정합에 활용이 가능하다.

Although a non-rigid registration has high demands in clinical practice, it has a high computational complexity and it is very difficult for ensuring the accuracy and robustness of registration. This study proposes a method of applying a non-rigid registration to 3D magnetic resonance images of brain in an unsupervised learning environment by using a deep-learning network. A feature vector between two images is produced through the network by receiving both images from two different patients as inputs and it transforms the target image to match the source image by creating a displacement vector field. The network is designed based on a U-Net shape so that feature vectors that consider all global and local differences between two images can be constructed when performing the registration. As a regularization term is added to a loss function, a transformation result similar to that of a real brain movement can be obtained after the application of trilinear interpolation. This method enables a non-rigid registration with a single-pass deformation by only receiving two arbitrary images as inputs through an unsupervised learning. Therefore, it can perform faster than other non-learning-based registration methods that require iterative optimization processes. Our experiment was performed with 3D magnetic resonance images of 50 human brains, and the measurement result of the dice similarity coefficient confirmed an approximately 16% similarity improvement by using our method after the registration. It also showed a similar performance compared with the non-learning-based method, with about 10,000 times speed increase. The proposed method can be used for non-rigid registration of various kinds of medical image data.

Deeplearning is applied to the field of natural language processing, and the learning model by encoder-decoder model is used. Neural network-based natural language processing techniques mostly use sequence-to-sequence models and train the model with supervised learning. The success of such end-to-end deep neural network depends on securing a large amount of learning data to train the model. It takes a lot of time and money to build a large I / O dataset. Recent studies are looking into ways to secure performance by expanding scarce datasets. In this paper, we describe the data set expansion methods, such as the denoising training method, transfer learning, and pre-learning BERT. We discuss the recent application of data set expansion to. the machine translation system

최근 스마트폰의 보급 및 웹 서비스의 도입으로 온라인 사용자들은 대규모의 콘텐츠를 시간과 장소에 관계없이 접할 수 있게 되었다. 그러나 사용자들은 대규모의 콘텐츠 사이에서 원하는 콘텐츠를 찾는 데 어려움을 겪게 되었다. 이러한 문제를 해결하기 위해 다양한 분야에서 사용자 모델링 및 추천 시스템에 대한 연구가 활발하게 수행되었다. 그러나 정보 환경의 변화에 따른 시니어 계층의 적극적인 변화에도 불구하고 시니어 계층에 초점을 맞춘 사용자 모델링 및 추천 시스템에 대한 연구는 매우 부족한 실정이다. 이에 본 논문에서는 기계 학습 방법을 기반으로 스마트 시니어 계층의 선호도를 파악할 수 있는 모델링 방법을 제안하고, 스마트 시니어 분류 모델을 개발 한다. 이 결과, 스마트 시니어 계층의 선호도를 파악할 수 있을 뿐만 아니라 스마트 시니어 분류 모델 개발을 통해 시니어 사용자에게 가장 적합한 활동 및 콘텐츠를 제공하는 콘텐츠 추천 연구에 대한 발판을 마련하였다.

With the recent spread of smartphones and the introduction of web services, online users can access large-scale content regardless of time or place. However, users have had trouble finding the content they wanted among large-scale content. To solve this problem, user modeling and content recommendation system have been actively studied in various fields. However, in spite of active changes in senior groups according to the changes in information environment, research on user modeling and content recommendation system focused on senior groups are insufficient. In this paper, we propose a method of modeling smart senior based on their preference, and further develop a smart senior classification model using machine learning methods. As a result, we can not only grasp the preferences of smart seniors, but also develop a smart senior classification model, which is the foundation for the research of a recommendation system which will provide the activities and contents most suitable for senior groups.

The objective of this study is to evaluate the mechanical behaviors and structural integrity of the weldment of high strength steel by using an acoustic emission (AE) techniques. Monotonic simple tension and AE tests were conducted against the 3 kinds of welded specimen. In order to analysis the effectiveness of weldability, joinability and structural integrity, we used K-means clustering method as a unsupervised learning pattern recognition algorithm for obtained multi-variate AE main data sets, such as AE counts, energy, amplitude, hits, risetime, duration, counts to peak and rms signals. Through the experimental results, the effectiveness of the proposed method is discussed.

Background Logistics service workers face elevated injury rates, with ankle-related incidents comprising 23% of workplace injuries. Traditional binary foot posture classifications may oversimplify the complex biomechanical relationships affecting dynamic balance performance in occupational settings. Purpose To apply unsupervised machine learning clustering to identify distinct foot posture phenotypes among logistics service workers and compare dynamic balance performance between identified clusters using the Y-Balance Test. Study design Cross-sectional observational study Methods A total of 112 logistics service workers were analyzed using K-means clustering based on age, body mass index, work duration, navicular drop, and resting calcaneal stance position. Silhouette score analysis determined optimal cluster number. Dynamic balance was assessed using the Y-Balance Test, measuring reach distances in anterior, posteromedial, and posterolateral directions. Results Four distinct phenotypes emerged: rearfoot valgus-dominant pronated (n=22), midfoot collapse-dominant pronated (n=32), age-related (n=23), and supinated (n=35) foot types. Significant differences in dynamic balance performance were observed in the posterolateral direction (F=3.900, p=0.011). The supinated phenotype demonstrated superior posterolateral reach performance (77.76±13.63%) compared to midfoot collapse-dominant (66.18±16.37%, p=0.003) and age-related phenotypes (67.78±14.97%, p=0.018). Conclusions Unsupervised machine learning successfully identified naturally occurring foot posture phenotypes with distinct dynamic balance characteristics. Midfoot collapse-dominant pronation demonstrated greater balance impairments than rearfoot valgus patterns, supporting the implementation of phenotype-specific interventions for workplace injury prevention in logistics workers.

PurposeThis study aimed to classify self-care patterns among Korean adults with prediabetes using an unsupervised machine learning approach. The classification was grounded in Orem’s Self-Care Theory, focusing on self-care demands, self-care agencies, and self-care behaviors. MethodsA secondary data analysis was conducted using the 2023 Korea National Health and Nutrition Examination Survey. Variables were selected and categorized according to the theoretical components of Orem’s model. Principal component analysis was applied for dimensionality reduction, followed by K-means clustering to identify distinct self-care pattern groups. All variables were standardized using min-max normalization. Group differences were examined using analysis of variance and the chi-square test. ResultsThree self-care pattern groups were identified: the high self-care performance group, the latent self-care risk group, and the self-care vulnerable group. These groups exhibited distinct profiles across self-care demands, agencies, and behaviors. Significant intergroup differences were also observed in education level, income, health literacy, fasting blood glucose, and hemoglobin A1c levels. ConclusionSelf-care patterns among adults with prediabetes can be effectively classified through unsupervised learning techniques. The findings highlight the importance of developing tailored nursing interventions that consider multidimensional self-care profiles. This study underscores the applicability of Orem’s Self-Care Theory and demonstrates the potential of machine learning in identifying at-risk subgroups for early intervention.

Background: The rapid growth of the aging population has increased the demand for technologies related to muscle health, assessment, prevention, and functional support. Purpose: This study aimed to analyze technology trends in muscle-related innovations using natural language processing and machine learning applied to South Korean patent and utility model data. Study design: Descriptive study using unsupervised machine learning. Methods: A total of 2,836 records were analyzed using combined title and abstract texts. Semantic embeddings were generated using Sentence-BERT, followed by dimensionality reduction with Uniform Manifold Approximation and Projection and clustering with Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). Keywords for each cluster were extracted using term frequency-inverse document frequency, and annual filing patterns were examined to identify the trends. Results: Two dominant macro-domains were identified, including a Wearable and Sensor Technology domain (n=1,695) and a Biological and Preventive Technology domain (n=903). HDBSCAN yielded substantially better cluster validity, with a silhouette coefficient of 0.610 and Davies-Bouldin index of 0.525. Within these macro-domains, sub-domains related to stem cell and regenerative therapy, functional food and bioactive extracts, antibody and antiviral therapeutics, aging and sarcopenia mechanisms, and massage devices were identified. Conclusions: These findings provide an overview of muscle-related technology development and demonstrate the value of machine learning-based patent text analysis for technology landscape mapping and innovation assessment.

Researchers and healthcare professionals have paid attention to artificial intelligence (AI) in the healthcare industry. The application of effective and efficient AI techniques can help to increase the precision of medical decision. AI offers great advances for countries currently struggling with complex healthcare systems and a physician shortage. In previous research in the field of health management, big data analytics in healthcare has been a significant subject from multidisciplinary aspects. To detect the onset of disease in the large-scale data, examining image and video sources and social media data is required. The literature represents various AI applications for healthcare services as well as an unexplored area of medical research emphasizes medical decision-making, patient diagnostics data, and network of health services. This study presents an innovative datadriven approach with knowledge-based analysis utilizing AI techniques. The proposed method develops a comprehensive approach including machine learning methods and its application to healthcare big data analysis. The significance of identifying and covering the key AI applications for healthcare is suggested. The result highlights that organization may greatly benefit from the use of this technology in healthcare operations with AI-based solutions to provide different treatment options and personalized therapies. With the applications of AI techniques, the general effectiveness of hospitals and healthcare systems may be increased, and overall healthcare costs may be reduced.

학교폭력은 사회적 문제로 인식되고 있으며 이를 예방하기 위한 다양한 노력도 함께 이루어지고 있다. 본 연 구에서는 학생들 상호 간의 친밀도, 대화 빈도 및 모둠 선호도 데이터를 분석함으로써 교우 관계를 파악하고 이 를 통해 궁극적으로 학교폭력을 예방할 수 있는 시스템을 개발하였다. 리커트 척도 설문을 활용하여 학급 내 학 생들과의 친밀도, 대화 빈도 및 모둠 선호도를 각각 수치화하였고, 설문 데이터는 K-means 알고리즘을 활용하 여 적정한 개수의 클러스터로 군집화 및 시각화하였다. 담임교사는 학급 내 학생들 간의 친밀도, 대화 빈도 및 모둠 선호도 분석 데이터를 그래프의 형태로 시각적으로 확인하고, 이를 근거로 특정 그룹의 학생 개별 상담 및 학급 운영 등 학교폭력 예방을 위한 참고 자료로 활용 가능하다. 데이터 분석 결과는 기존에 교사가 학생 관찰 등의 방법으로 정성적으로 파악하고 있던 교우 관계와 상당 부분 일치하였고, 이는 본 데이터 분석 시스템이 담 임교사의 학급 내 교우 관계 파악을 위한 정량적 근거 자료로 유의미하게 활용될 수 있음을 의미한다. 한계점은 학생들의 주관적인 기준으로 인해 설문 결과가 왜곡될 수 있는 점이다. 본 연구가 담임교사의 학급 내 교우 관 계 파악 및 학교폭력 예방 노력에 실질적인 도움을 제공하며 학교폭력 예방에 기여하기를 기대한다.

School violence has long been recognized as a social problem, and various efforts have been made to prevent it. In this study, we propose a system that can prevent school violence by analyzing data on the frequency of conversations between students, friendship and preference to be in the same group. This data was quantified using a Likert scale questionnaire, and also grouped into the appropriate number of clusters using the K-means algorithm. Additionally, the homeroom teacher observed the frequency and nature of conversations between students, and targeted specific individuals or groups for counseling and intervention, with the aim of reducing school violence. Data analysis revealed that the teachers’ qualitative observations were consistent with the quantified data based on student questionnaires, and therefore applicable as quantitative data towards the identification and understanding of student relationships within the classroom. The study has potential limitations. The data used is subjective and based on peer evaluations which can be inconsistent as the students may use different criteria to evaluate one another. It is expected that this study will help homeroom teachers in their efforts to prevent school violence by understanding the relationships between students within the classroom.

본 연구는 비교사학습의 대표적인 방법 중 하나인 코호넨의 자기조직화 방법을 기반으로 BCI(Brain-Computer Interface)에 적용 가능한 자율적 기계학습방법을 제안한다. 이를 위해 상호작용 함수를 이용한 학습영역조정방법과 자율적 기계학습규칙을 제안하였다. 학습영역조정과 기계학습은 코호넨의 자기조직화 방법을 기반으로 한 상호작용 함수에 의한 측면제어효과를 이용하였다. 승자 뉴런을 결정하고 난 후 학습 규칙에 따라 뉴런의 연결강도를 조정하 고 학습 횟수가 증가함에 따라 학습영역이 점차 감소하여 출력층 뉴런 가중치들의 입력을 향한 유동을 완화시켜 네 트워크가 평형 상태(equilibrium state)에 도달하여 학습을 마칠 수 있는 자율적 기계학습을 제안하였다.

This paper proposes an autonomous machine learning method applicable to the BCI(Brain-Computer Interface) is based on the self-organizing Kohonen method, one of the exemplary method of unsupervised learning. In addition we propose control method of learning region and self machine learning rule using an interactive function. The learning region control and machine learning was used to control the side effects caused by interaction function that is based on the self-organizing Kohonen method. After determining the winner neuron, we decided to adjust the connection weights based on the learning rules, and learning region is gradually decreased as the number of learning is increased by the learning. So we proposed the autonomous machine learning to reach to the network equilibrium state by reducing the flow toward the input to weights of output layer neurons.

최근 차량 내부와 외부 네트워크를 연결하는 차량 내 네트워크 기술이 발전함에 따라 CAN(Controller Area Network)을 통해 다양한 정보를 실시간으로 교환할 수 있다. 하지만 CAN 네트워크는 사이버 공격에 취약 하다. 따라서 본 연구에서는 지도 및 비지도 딥러닝 모델 기반 침입 탐지 방법을 제안한다. 실험 데이터로는 CAN ID로 구성된 시계열 데이터를 사용하였다. 지도학습 기반 침입 탐지 방법은 전반적으로 높은 탐지 성능을 보였다. 정상 데이터로만 학습하여 탐지하는 비지도 기반 방식의 경우에도 CAN에 오작동을 일으킬 수 있는 새로운 공격 유형(Spoofing, DoS)에 대한 탐지율이 높았다. 비지도학습을 위해 LSTM(Long Short-Term Memory) 기반 오 토인코더와 IF(Isolation Forest)를 사용하였다. 따라서 제안한 기법은 알려진 공격 외에 새로운 공격을 탐지하는 데도 효과적인 것을 확인하였다.

Recently, with the development of in-vehicle network technology that connects the internal and external networks of vehicle, various information can be exchanged in real time through CAN (Controller Area Network). However, the CAN networks are vulnerable to cyberattacks. Therefore, in this study, we propose an intrusion detection method based on supervised and unsupervised deep learning models. Time series data consisting of CAN IDs were used as experimental data. The supervised learning-based intrusion detection method showed high overall detection performance. Even in the case of an unsupervised method that learns and detects only normal data, the detection rate for new attack types (Spoofing, DoS) that can cause malfunctions in CAN was high. For unsupervised learning, an LSTM (Long Short-Term Memory) based Autoencoder and IF (Isolation Forest) were used. Therefore, it was confirmed that the proposed technique is effective in detecting new attacks in addition to known attacks.

1인 창조기업에 대한 정부의 지원이 확대되고, 인공지능과 IT 기술의 발달로 1인 창조기업의 성장할 수 있는 환경이 좋아졌다. 이로 인해 1인 창조기업에 대한 사람들의 관심이 증가하고, 1인 창조기업 지원센터를 찾는 청년 창업가의 숫자도 크게 증가하고 있다. 본 연구는 비지도 기계학습과 베이지안 네트워크를 결합한 새로운 유형의 확률적 구조방정식 모형, 즉 PSEM을 사용하여 국내 1인 창조기업의 당기순이익에 영향을 주는 요인을 분석하고자 한다. 아울러 해당 영향요인들이 국내 1인 창조기업의 당기순이익에 미치는 인과관계 값을 산출하기 위하여 What-If 분석을 실시하였다. 분석 결과 당기 순이익의 변화에 큰 영향을 받지 않는 요인들은 서비스 업종일 때 ‘초기 자본금’, ‘창업자특성’, ‘준비 기간’, ‘창업자 배경’, ‘회사 특성’으로 나타났다. 서비스 업종을 제외하였을 때는 ‘창업자 배경’, ‘인식’, ‘초기 자본금’, ‘회사 특성’, ‘창업자특성’으로 분석되었다. 당기순이익에 영향을 주는 요인은 ‘수익’, ‘종업원’, ‘판매’임을 발견하였다.

In the era of economic recession caused by covid-19 pandemic, government expanded various types of supports for one-man creative companies with a bid to promoting healthier ecosystem for start-ups. Furthermore, the 4th industrial revolution atmospheres emboldened by AI technologies have countenanced young entrepreneurs’attempts to launch their own one-man creative companies. However, despite positive trends with one-man creative companies like this, few studies exist investigating which factors affect its net profit. To fill the research void, this study proposes a novel method of probabilistic structural equation model (PSEM) combining unsupervised machine learning and Bayesian networks. Empirical results revealed that those influential factors affecting one-man creative companies’ net profit encompass ‘income’, ‘number of employees’, and ‘sales’. On the contrary, multiple factors initially deemed to be significant to the net profit were excluded, like “background of the CEO”, “characteristics of the company”, “initial fund”, ‘preparation’, and “characteristics of the CEO” in service company. The others company’s results showed that ‘background of the CEO’, ‘perception’, ‘initial fund’, ‘characteristics of the company’, and ‘characteristics of the CEO’ were not included in factors affected by changes of net profit.

인공 위성을 통해 촬영된 해수면 위성 영상 정보를 실제 응용 분야에 활용하기 위해서는 손실된 온도 정보를 인위적 으로 복원하는 기술이 필수적이다. 본 논문에서는 해수면 온도 위성영상 복원을 위해 정답 영상이 필요 없는 비지도 (Unsupervised) 기반 외부-내부 학습으로 구성된 딥 러닝 모델을 제안한다. 제안 방법에서는 손실된 해수면 온도 영상의 정답 영상이 부재해도 손실된 해수면 온도 정보를 신뢰성 있게 복원할 수 있게 한다. 제안 복원 방법은 외부- 내부 학습 2단계의 과정으로 구성되어 있으며 1단계 외부 학습 과정에서는 비지도 학습을 기반으로 손실된 영역의 구조적 복원을 진행한다. 2단계 내부 학습 과정에서는 구조적 복원 결과와 손실 영상을 기반으로 색상 복원을 진행 한다. 제안 비지도 학습 기반 복원 딥 러닝 모델은 최신 다른 비지도 학습 모델들과 비교할 때 우수한 복원 성능을 달성했으며 정답 영상을 필연적으로 활용하는 지도 학습 기반 복원 모델들과 비교해도 유사한 복원 정확도를 보였 다. 또한 실제 측정한 해수면 온도 값과 비교하여 평균적으로 섭씨 1.8도 오차내에 복원 정확도를 달성하였다.

In order to utilize Sea Surface Temperature (SST) satellite image in practical applications, the techniques for restoring the lost temperature information is essential. We proposed a so-called unsupervised external-internal learning (UNEIL) model that does not require ground-truth images during the restoration of SST satellite images. Using the proposed UNEIL, it is possible to restore sea surface temperature information lost due to occluded clouds without using ground-truth images and excellent quality restoration results can be obtained. The proposed UNEIL largely consists of two-step learning steps: external and internal learning. In external learning, structural information of the lost area is restored in an unsupervised learning way. Next, internal learning based color restoration is performed based on structural restoration results. Experimental results showed that our UNEIL outperforms other state-of-the-art unsupervised restoration methods in terms of restoration accuracy and also achieves quite similar restoration accuracy, even compared to supervised restoration models that inevitably utilizes ground-truth images. The restoration accuracy of our UNEIL is achieved within an error of 1.8 degree Celsius on average compared to the actual sea surface temperature value.

Companies are building and utilizing their own data analysis systems according to business characteristics in the distributed cloud. However, as businesses and data types become more complex and diverse, the demand for more efficient analytics has increased. In response to these demands, in this paper, we propose an unsupervised learning-based data analysis agent to which reinforcement learning is applied for effective data analysis. The proposal agent consists of reinforcement learning processing manager and unsupervised learning manager modules. These two modules configure an agent with k-means clustering on multiple nodes and then perform distributed training on multiple data sets. This enables data analysis in a relatively short time compared to conventional systems that perform analysis of large-scale data in one batch.

Recently, as businesses and data types become more complex and diverse, efficient data analysis using machine learning is required. However, since communication in the cloud environment is greatly affected by network latency, data analysis is not smooth if information delay occurs. In this paper, SPT (Safe Proper Time) was applied to the cluster-based machine learning data analysis agent proposed in previous studies to solve this delay problem. SPT is a method of remotely and directly accessing memory to a cluster that processes data between layers, effectively improving data transfer speed and ensuring timeliness and reliability of data transfer.