http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
머신러닝을 활용한 응급실 내원 환자 퇴실 조치 결정 조기 예측
황하은,강현구,이의선,김정윤,윤영훈,김성범 대한산업공학회 2021 대한산업공학회지 Vol.47 No.3
Overcrowding within emergency departments (ED) affects patient satisfaction and quality of care. The leading causes of ED overcrowding are systematic delays between procedures and patient disposition after ED treatment. Early prediction of patient disposition can improve patient flow and optimize allocation of hospital resources. While studies for predicting disposition using machine learning methods have been actively conducted abroad, few have been conducted in South Korea in spite of the lagging emergency medical environment. Previous studies are limited to binary predictions; either hospital admission or discharge. In this study, we attempted to predict disposition (discharge, general ward admission, ICU admission) of patients using initial information of ED patients from the Korean national emergency department information system (NEDIS). We used five machine learning methods including logistic regression, decision tree, random forest, CatBoost, and TabNet. The results showed that CatBoost yielded the best performance. This result can aid in decision making by providing standard indicators for hospital admission.
피카소 작품과 해체주의 특성을 활용한 업사이클 패션디자인 개발
황하은,이연희 복식문화학회 2023 服飾文化硏究 Vol.31 No.6
This study aims to merge Picasso’s expressive elements and deconstructive fashion’s formative traits, proposing an upcycle fashion design that fuses artistic and philosophical aspects. The analysis of Picasso’s Cubism identified qualities like liberating revolution, fluidity of vision, geometric reducibility, complex symbolism, and creative imitation. The analysis of Derrida’s deconstructionism revealed expressive traits: uncertainty, intertextuality, différance, and dis-de phenomenon. An upscale fashion design was developed based on six Picasso works featuring women. The design was created using the fashion design software CLO 3D and integrated clothing waste and scrap fabrics as materials.The results are as follows. First, upcycle fashion was viewed from a new perspective based on Picasso and Derrida’s values. This perspective suggested creating better ethical values by upholding environmental protection in novel ways that overcome limitations rather than destroy existing values indiscriminately. Second, upcycle fashion design methodologies were derived from various perspectives utilizing formative features of Picasso’s works and specific expressive features of deconstructed fashion. Third, the direction of mitigating waste and pollution from clothing production and transportation was revealed by making clothes in a virtual space using the CLO 3D program. This study contributed to obtaining various methods for developing upcycle fashion designs using own methods of Picasso and Derrida to diversify the approaches of upcycling, which is relatively stagnant in disassembling.
머신러닝 및 베이지안 최적화를 이용한 타이어 최적 설계
황하은,조윤상,황석철,김성범 대한산업공학회 2022 대한산업공학회지 Vol.48 No.4
Product design optimization plays an important role in the manufacturing industry. In the tire manufacturing industry, design optimization process traditionally involves generation of tire design candidates and quality prediction by using finite element analysis (FEA). However, this traditional process requires expert’s experiences to derive design candidates that satisfies target quality. In addition, FEA requires a lot of time to obtain the prediction results although it provides accurate predictive performance. To overcome these issues, we propose Bayesian optimization based on a predictive model for the tire design. We train a model that can predict multiple quality variables and perform Bayesian optimization that can optimize numerical and categorical variables simultaneously. Results show that the proposed method can effectively predict and optimize the tire design with reduced time complexity.
머신러닝을 활용한 응급실 내원 환자 퇴실 조치 결정 조기 예측
황하은(Haeun Hwang),김성범(Seoung Bum Kim) 대한산업공학회 2020 대한산업공학회 추계학술대회논문집 Vol.2020 No.11
Overcrowding within emergency departments (ED) affects patient satisfaction and quality of care. One of the leading causes of ED overcrowding is the boarding of hospitalized patients in the ED as they await bed placement. Early prediction of disposition of patients can improve patient flow and optimize allocation of hospital resources and bed. Prediction of disposition using supervised machine learning methods are being actively researched abroad. However, there is a need for research suitable for the emergency medical environment in Korea. Previous studies were generally limited to predictions for disposition of either hospital admission or discharge. In this study we attempted to predict disposition (Discharge, General ward admission, ICU admission, Transfer) of patients using initial information of ED patients from the Korean National Emergency Department Information System (NEDIS). We used light gradient boosted machines, Catboost and TabNet. The results showed that TabNet yielded the best performance. This result can aid in decision making by providing standard indicators for hospital admission.
Feature Extraction and Deep Learning Model for Respiratory Sound Analysis
정기원,황하은,김성범 한국품질경영학회 2021 한국품질경영학회 학술대회 Vol.2021 No.-
In the medical field, doctors diagnose respiratory disease by auscultating the patient’s respiratory sound. It means that the subjective judgment of the doctor is diagnosed with respiratory diseases rather than the quantitative assessment method. However, the subjective diagnosis method relies on the experience of the doctor and can lead t misdiagnosed results. To address these issues, it is important to derive quantitative indicators based on the analysis of respiratory sound data ad utilize it as an objective aid for the diagnosis of the doctor. In this study, we propose using a Hierarchical Attention Network (HAN) model for respiratory sound analysis. This method can reflect hierarchical patterns of respiratory sounds that consist of time and frequency domain and allows doctors to interpret the important feature of respiratory sounds. In addition, we propose a feature extraction method that applies several features of respiratory sound as stacked channels. We conducted experiments on real-world respiratory sound data to demonstrate the effectiveness and applicability of our method. The experimental results showed that the proposed method outperformed the existing methods for respiratory sound analysis. We believe that the proposed method can contribute to diagnosis in the medical field and various industries where the interpretation of sound data is important.