• 한국산업정보학회논문지
• /
• 제27권5호
• /
• pp.1-11
• /
• 2022

딥러닝 기술의 발전에 따라 학습을 통해 선호도 랭킹 추정을 하기 위한 다양한 연구 개발이 진행되고 있으며, 웹 검색, 유전자 분류, 추천 시스템, 이미지 검색 등 여러 분야에 걸쳐 이용되고 있다. 딥러닝 기반의 선호도 랭킹을 추정하기 위해 근사(approximation) 알고리즘을 이용하는데, 이 근사 알고리즘에서 적정한 정도의 정확도를 보장할 수 있도록 모든 비교 대상에 k번 이상의 비교셋을 구축하게 되며, 어떻게 비교셋을 구축하느냐가 학습에 영향을 끼치게 된다. 이 논문에서는 크라우드 소싱 기반의 딥러닝 선호도 측정을 위한 쌍체 비교 셋을 생성하는 새로운 알고리즘인 k-disjoint 비교셋 생성 알고리즘과 k-체이닝 비교셋 생성 알고리즘을 제안한다. 특히 k-체이닝 알고리즘은 기존의 원형 생성 알고리즘과 같이 데이터 간의 연결성을 보장하면서도 안정적인 선호도 평가를 지원할 수 있는 랜덤적 성격도 함께 가지고 있음을 실험에서 확인하였다.

• 한국인터넷방송통신학회논문지
• /
• 제22권5호
• /
• pp.177-183
• /
• 2022

최근 COVID-19로 인해 홈 트레이닝의 관심도가 증가하고 있다. 이에 따라 HAR(human activity recognition) 기술을 홈 트레이닝에 적용한 연구가 진행되고 있다. 기존 HAR 분야의 논문에서는 동적인 자세보다는 앉기, 일어서기와 같은 정적인 자세들을 분석한다. 본 논문은 동적인 운동 자세를 분석하여 사용자의 운동 자세 정확도를 보여주는 딥러닝 모델을 제안한다. AI hub의 피트니스 이미지를 blaze pose를 사용하여 사람의 자세 데이터를 분석한다. 3개의 딥러닝 모델: RNN(recurrnet neural networks), LSTM(long short-term memory networks), CNN(convolution neural networks)에 대하여 실험을 진행한다. RNN, LSTM, CNN 모델의 f1-score는 각각 0.49, 0.87, 0.98로 CNN 모델이 가장 적합하다는 것을 확인하였다. 이후 연구로는, 다양한 학습 데이터를 사용하여 더 많은 운동 자세를 분석할 예정이다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• 아태비즈니스연구
• /
• 제13권3호
• /
• pp.449-461
• /
• 2022

Purpose - The purpose of this study to explore factors consumers prioritize when purchasing seafood online. The originality of the study lies on adopting AHP-based approach in analyzing prioritized purchasing factors of seafood online. Design/methodology/approach - A survey was conducted targeting Korean consumers who have purchased seafood online. AHP method was applied to rank factors consumers prioritize before making decision. Findings - First, product's factor ranked first among other high level factors including delivery service, seller, online platform. Second, sanitation, taste, country of origin ranked first, second, third respectively, within product's factors. Third, safe delivery, timeliness, information accuracy ranked first, second, third respectively, within delivery factors. Fourth, consumer reviews, consumer response ability, promotion ranked first, second, third within seller factors. Fifth, Personal information management system, credibility, user-friendliness ranked first, second, third, within online platform factors. Research implications or Originality - To activate seafood online market, it is crucial to assure consumers that the seafood is well managed in a sanitary way from the production site to table. Existing government programs such as seafood traceability system, HACCP, and cold-chain infrastructure needs improvement. Due to highly perishable characteristic of seafood, delivery factors matter when purchasing online. Online platforms needs to continue to improve delivery service. Seafood products are mostly not branded and without objective information about their properties. Creating quality classification and seafood brands are likely to help consumers chose seafood online.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권9호
• /
• pp.2991-3007
• /
• 2022

Two dimensional locality preserving projections (2D-LPP) is an improved algorithm of 2D image to solve the small sample size (SSS) problems which locality preserving projections (LPP) meets. It's able to find the low dimension manifold mapping that not only preserves local information but also detects manifold embedded in original data spaces. However, 2D-LPP is simple and elegant. So, inspired by the comparison experiments between two dimensional linear discriminant analysis (2D-LDA) and linear discriminant analysis (LDA) which indicated that matrix based methods don't always perform better even when training samples are limited, we surmise 2D-LPP may meet the same limitation as 2D-LDA and propose a novel matrix exponential method to enhance the performance of 2D-LPP. 2D-MELPP is equivalent to employing distance diffusion mapping to transform original images into a new space, and margins between labels are broadened, which is beneficial for solving classification problems. Nonetheless, the computational time complexity of 2D-MELPP is extremely high. In this paper, we replace some of matrix multiplications with multiple multiplications to save the memory cost and provide an efficient way for solving 2D-MELPP. We test it on public databases: random 3D data set, ORL, AR face database and Polyu Palmprint database and compare it with other 2D methods like 2D-LDA, 2D-LPP and 1D methods like LPP and exponential locality preserving projections (ELPP), finding it outperforms than others in recognition accuracy. We also compare different dimensions of projection vector and record the cost time on the ORL, AR face database and Polyu Palmprint database. The experiment results above proves that our advanced algorithm has a better performance on 3 independent public databases.

• Physical Therapy Rehabilitation Science
• /
• 제11권3호
• /
• pp.285-295
• /
• 2022

Objective: The purpose of this study is to use logistic regression and decision tree analysis to identify the factors that affect the success or failurein the national physical therapy examination; and to build and compare predictive models. Design: Secondary data analysis study Methods: We analyzed 76,727 subjects from the physical therapy national examination data provided by the Korea Health Personnel Licensing Examination Institute. The target variable was pass or fail, and the input variables were gender, age, graduation status, and examination area. Frequency analysis, chi-square test, binary logistic regression, and decision tree analysis were performed on the data. Results: In the logistic regression analysis, subjects in their 20s (Odds ratio, OR=1, reference), expected to graduate (OR=13.616, p<0.001) and from the examination area of Jeju-do (OR=3.135, p<0.001), had a high probability of passing. In the decision tree, the predictive factors for passing result had the greatest influence in the order of graduation status (x²=12366.843, p<0.001) and examination area (x²=312.446, p<0.001). Logistic regression analysis showed a specificity of 39.6% and sensitivity of 95.5%; while decision tree analysis showed a specificity of 45.8% and sensitivity of 94.7%. In classification accuracy, logistic regression and decision tree analysis showed 87.6% and 88.0% prediction, respectively. Conclusions: Both logistic regression and decision tree analysis were adequate to explain the predictive model. Additionally, whether actual test takers passed the national physical therapy examination could be determined, by applying the constructed prediction model and prediction rate.

• 반도체디스플레이기술학회지
• /
• 제20권4호
• /
• pp.161-166
• /
• 2021

Errors in the semiconductor process are generated by a change in the state of the equipment, and errors usually arise when the state of the equipment changes or when parts that make up the equipment have flaws. In this investigation, we anticipated that aging of the mass flow controller in the plasma enhanced chemical vapor deposition SiO2 thin film deposition method caused a minute flow rate shift. In seven cases, fourier transformation infrared film quality analysis of the deposited thin film was used to characterize normal and pathological processes. The plasma condition was monitored using optical emission spectrometry data as the flow rate changed during the procedure. Preprocessing was used to apply the collected OES data to the artificial immune system algorithm, which was then used to process diagnosis. Through comparisons between datasets, the learning algorithm compared classification accuracy and improved the method. It has been confirmed that data characterized as a normal process and abnormal processes with differing flow rates may be discriminated by themselves using the artificial immune system data mining method.

• 분석과학
• /
• 제35권2호
• /
• pp.69-81
• /
• 2022

The method of measuring and classifying the energy category of neutrons directly using raw data acquired through a CZT detector is not satisfactory, in terms of accuracy and efficiency, because of its poor energy resolution and low measurement efficiency. Moreover, this method of measuring and analyzing the characteristics of low-energy or low-activity gamma-ray sources might be not accurate and efficient in the case of neutrons because of various factors, such as the noise of the CZT detector itself and the influence of environmental radiation. We have therefore developed an efficient method of analyzing radiation characteristics using a neutron and gamma-ray analysis algorithm for the rapid and clear identification of the type, energy, and radioactivity of gamma-ray sources as well as the detection and classification of the energy category (fast or thermal neutrons) of neutron sources, employing raw data acquired through a CZT detector. The neutron analysis algorithm is based on the fact that in the energy-spectrum channel of 558.6 keV emitted in the nuclear reaction ¹¹³Cd + ¹n → ¹¹⁴Cd + in the CZT detector, there is a notable difference in detection information between a CZT detector without a PE modulator and a CZT detector with a PE modulator, but there is no significant difference between the two detectors in other energy-spectrum channels. In addition, the gamma-ray analysis algorithm uses the difference in the detection information of the CZT detector between the unique characteristic energy-spectrum channel of a gamma-ray source and other channels. This efficient method of analyzing radiation characteristics is expected to be useful for the rapid radiation detection and accurate information collection on radiation sources, which are required to minimize radiation damage and manage accidents in national disaster situations, such as large-scale radioactivity leak accidents at nuclear power plants or nuclear material handling facilities.

• 한국정보통신학회논문지
• /
• 제26권5호
• /
• pp.755-762
• /
• 2022

본 논문에서는 학습 예측이 가능한 군집적 알고리즘으로 COVID-19에서 상황인식정보인 질병의 속성정보와 클러스터링를 이용한 군집적 알고리즘을 제안한다. 클러스터링 내에서 처리되는 군집 데이터는 신규 또는 새롭게 입력되는 정보가 상호관계를 예측하기 위해 분류 제공되는데, 이때 새롭게 입력되는 정보가 비교정보에서 오염된 정보로 처리되면 기존 분류된 군집으로부터 벗어나게 되어 군집성을 저하시키는 요인으로 작용하게 된다. 본 논문에서는 COVID-19에서의 질병속성 정보내 K-means알고리즘을 이용함에 있어 이러한 문제를 해결하기 위해 질병 상호관계 정보 추출이 가능한 사용자 군집 분석 방식을 제안하고자 한다. 제안하는 알고리즘은 자율적인 사용자 군집 특징의 상호관계를 분석학습하고 이를 통하여 사용자 질병속성간에 따른 클러스터를 구성해 사용자의 누적 정보로부터 클러스터의 중심점을 제공하게 된다. 논문에서 제안된 COVID-19의 다중질병 속성정보군집단위로 분류하고 학습하는 알고리즘은 적용한 모의실험 결과를 통해 사용자 관리 시스템의 예측정확도가 학습과정에서 향상됨을 보여주었다.

• 한국정보통신학회논문지
• /
• 제26권5호
• /
• pp.675-681
• /
• 2022

최근 IoT 기술과 AI의 발전에 따라 다양한 분야에서 무인화와 자동화가 진행되고 있으며, 사물인식과 객체분류 등 자동화의 기반이 되는 영상처리에 대한 관심이 높아지고 있다. 영상처리 과정에서 잡음 제거는 영상의 품질 또는 시스템의 정확성과 신뢰성에 큰 영향을 미치는 과정으로 다양한 연구가 진행되고 있으나, 영상에서 임펄스 잡음의 밀도가 높은 영역에 대한 영상을 복원하기 어렵다는 문제점이 있다. 따라서 본 논문은 영상에서 임펄스 잡음 훼손된 영역을 복원하기 위해 부분 마스크와 라그랑지 보간법에 기반한 필터 알고리즘을 제안한다. 제안한 알고리즘은 필터링 마스크와 잡음 추정치를 서로 비교하여 필터링 과정을 스위칭하였으며, 영상의 저주파 및 고주파 성분에 따라 퍼지 가중치를 계산하여 영상을 복원하였다.