포항공대 분자 모델링 시스템 개발 및 생화학 분자 내에서의 Zn(II)의 배위에 관한 이론적 연구

윤장우 포항공과대학 1992 국내석사

포항공대 분자그래픽스 시스템 개발 및 물분자 집합체에 관한 이론적 연구

김성욱 포항공과대학 1990 국내석사

Renal Dysfunction and Occipital Cerebral Microbleeds in Posterior Reversible Encephalopathy Syndrome: Associations With Lesion Patterns and Hemorrhagic Complications

정동영 포항공과대학교 융합대학원 2026 국내박사

Background: Posterior reversible encephalopathy syndrome (PRES) is a clinico-radiological syndrome characterized by cerebral blood flow dysregulation and disruption of the blood–brain barrier (BBB). Intracranial hemorrhage (ICH) is a well-recognized complication associated with poor prognosis in PRES, and renal dysfunction has been proposed as a systemic risk factor related to BBB vulnerability. In addition, cerebral microbleeds (CMBs) have emerged as an imaging marker of endothelial injury. However, the interrelationships among renal dysfunction, lesion distribution and volume, CMBs, and ICH in PRES have not been fully elucidated. Methods: This retrospective observational study included patients diagnosed with PRES and comprised two complementary analyses. First, renal dysfunction was defined as an estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73 m² at symptom onset. PRES lesion volume and distribution were quantitatively measured using fluid-attenuated inversion recovery (FLAIR) imaging. Lesions were classified into typical regions (parietal–occipital–temporal lobes) and atypical regions. Logistic regression and multivariable regression analyses were performed for DMED 정동영, Dongyoung Jeong 20232585 Renal Dysfunction and Occipital Cerebral Microbleeds in Posterior Reversible Encephalopathy Syndrome: Associations With Lesion Patterns and Hemorrhagic Complications 가역적 후두뇌병증 증후군에서 신기능 저하와 후두엽 뇌미세출혈: 병변 양상 및 출혈 합병증과의 관련성 Medical Science and Engineering Program, 2026, 64P, Advisor: Chulhong Kim, Text in English statistical evaluation. Second, patients were categorized according to the presence of ICH on imaging obtained on the day of symptom onset. CMBs and ICH were assessed using gradient echo (GRE), susceptibility- weighted imaging (SWI), and computed tomography (CT). The brain was divided into nine regions, and the regional distribution of CMBs, including the occipital lobe, was analyzed. Results: The renal dysfunction analysis included 200 patients, among whom 94 (47%) had renal dysfunction. Patients with renal dysfunction had a significantly larger total lesion volume compared with those without renal dysfunction (144.7±125.2 cc vs. 110.5±93.2 cc; p=0.032), with a notably greater lesion volume in atypical regions (49.2±65.0 cc vs. 29.2±44.3 cc; p=0.013). A decrease in eGFR was independently associated with an increase in total lesion volume. However, no significant difference was observed in lesion reversibility between patients with and without renal dysfunction (35.2±67.5 cc vs. 18.8±33.4 cc; p=0.129). A total of 165 patients with PRES were included in the ICH analysis, of whom 19 (11.5%) presented with ICH. Patients with PRES complicated by ICH showed a significantly higher prevalence of CMBs in the parietal and occipital lobes. In particular, the presence of occipital CMBs was strongly associated with the occurrence of ICH (odds ratio 10.71, 95% confidence interval 1.44–79.63, p=0.020). Conclusions: This study demonstrates that renal dysfunction and occipital CMBs are closely associated with lesion severity and hemorrhagic complications in PRES. Renal dysfunction is linked to more extensive lesion formation involving atypical regions, while occipital CMBs may reflect localized endothelial vulnerability and increase the risk of ICH. These findings suggest an interaction between systemic factors and focal vascular injury in the pathophysiology of PRES and highlight the clinical importance of imaging-based risk assessment.

Quantifying the Divergence of Socio-Hydrological Responses to Extreme Rainfall: A Spatiotemporal Analysis of Urban Pluvial Floods

최기훈 포항공과대학교 융합대학원 2026 국내석사

Major urban inundation events are not merely hydrometeorological phenomena but complex socio-political events that transform how citizens perceive risk. As climate change intensifies flood volatility in East Asia, understanding the social dimensions of these disasters is critical. This study investigates "socio-hydrological coupling" by analyzing a unique dataset of administrative civil complaints (Minwon) in South Korea, distinct from commonly used social media data. We conducted a comparative analysis of two contrasting events: the August 2022 Seoul metropolitan inundation and the July 2023 central-southern inundation. Despite comparable precipitation intensities, the two events elicited fundamentally different social responses. First, a strong "socio-hydrological synchronization" was observed, with complaints surging in tandem with rainfall peaks (1-2 day lag), demonstrating the utility of administrative data as a high-fidelity social sensor. Second, the emotional landscapes diverged significantly: the 2022 event was dominated by 'Fear', while the 2023 event shifted toward 'Mistrust' and 'Outrage'. This presents empirical evidence supporting Sandman’s risk framework ("Risk = Hazard + Outrage") in an urban flood context, suggesting that public sentiment is modulated by the attribution of accountability rather than rainfall magnitude alone. Third, spatial analysis revealed a "decoupling" in 2023, where outrage clustered in jurisdictions associated with governance failure rather than physical damage. Ultimately, this study demonstrates the value of civil complaints not merely as grievances but as an "institutional footprint" for monitoring the socio-hydrological contract. It contributes to the literature by extending socio-hydrological concepts to the domain of administrative informatics, offering a methodological blueprint for resilient governance in the era of climate complexity. Keywords Urban Pluvial Flood, Extreme Rainfall, Socio-hydrological Coupling, Civil Complaints (Minwon), Emotion Analysis, Attribution Theory 극한 호우로 촉발된 대규모 도시 침수(urban inundation)는 단순한 수문기상학적 현상을 넘어, 시민들이 위험을 인식하는 방식을 근본적으로 변화시키는 복합적인 사회·정치적 사건이다. 기후 변화로 인해 동아시아 지역의 홍수 변동성이 심화됨에 따라, 이러한 재난의 사회적 차원을 이해하는 것은 매우 중요하다. 이에 본 연구는 기존의 소셜 미디어 데이터와 구별되는 한국의 독자적인 행정 민원(civil complaint) 데이터셋을 분석하여 '수문-사회 결합(socio- hydrological coupling)' 현상을 규명하였다. 본 연구는 2022 년 8 월 수도권 침수(자연재해 성격 우세)와 2023 년 7 월 중남부 침수(거버넌스 실패 성격 우세)라는 대조적인 두 사건을 비교 분석하였다. 유사한 강수 강도에도 불구하고, 두 사건은 근본적으로 다른 사회적 반응을 유발하였다. 첫째, 강수량 피크와 민원 급증 사이에 강력한 '수문-사회 동기화(socio-hydrological synchronization)' 현상(1~2 일 시차)이 관찰되었으며, 이는 행정 데이터가 고충실도(high-fidelity)의 사회적 센서로서 유용함을 입증한다. 둘째, 대중의 감정 지형(emotional landscape)은 현저하게 분화되었다. 2022 년 사건은 '공포(Fear)'가 지배적이었던 반면, 2023 년 사건은 '불신(Mistrust)'과 '분노(Outrage)'로 전이되었다. 이는 도시 홍수 맥락에서 샌드만(Sandman)의 위험 프레임워크("위험 = 유해성 + 분노")를 실증적으로 지지하며, 대중의 정서가 단순한 강우량이 아닌 책임 소재의 귀인(attribution)에 의해 조절됨을 시사한다. 셋째, 공간 분석 결과 2023 년에는 '비동조화(decoupling)' 현상이 확인되었으며, 분노는 물리적 피해 지역보다는 거버넌스 실패와 관련된 관할 구역에 집중되는 경향을 보였다. 결론적으로 본 연구는 민원을 단순한 불만 토로가 아닌, 수문-사회 계약을 모니터링하는 '제도적 발자국(institutional footprint)'으로서의 가치를 조명한다. 이는 사회수문학적 개념을 행정 정보학 영역으로 확장하여 학문적으로 기여하며, 기후 복잡성 시대에 회복탄력적 거버넌스를 구축하기 위한 방법론적 청사진을 제공한다.

감성분석 기반 7 가지 감정을 반영한 재구매 예측 및 고객 구매행동 분석

박해균 포항공과대학교 융합대학원 2024 국내석사

최근 들어 온라인 플랫폼의 발달로 인해 디지털 커머스(digital commerce)가 크게 성장하고 있다. 이에 따라, 소비자들이 주로 온라인 플랫폼을 통한 구매의사결정을 하면서 온라인 구매리뷰의 중요성이 기업과 소비자 측면에서 대두되고 있는 추세에 있다. 이와 관련하여, 기존의 CSA(Customer Sentimental Analysis)와 재구매예측 선행연구들이 많이 진행되었으나, 각각 긍부정의 단순한 2가지 감정에 대한 분석과 텍스트 데이터나 그 외의 정형 데이터 중 일부만을 활용하는 한계점을 보여왔다. 이를 보완하고자 본 연구에서는 Phase 1의 7가지 감정분류 모델에서 7가지 감정인 분노(Anger), 혐오(Disgust), 두려움(Fear), 기쁨(Happiness), 슬픔(Sadness), 놀람(Surprise), 중립(Neutrality)을 분류하도록 학습하였고 93%의 정확도를 보였다. 이어진 Phase 2의 재구매예측 모델에서는 리뷰 데이터에서 Phase 1을 통해 추출한 7가지 감정을 반영하였고, 감정을 반영하지 않은 모델의 정확도인 89.04%에 비해, 90.25%로 상대적으로 높은 성능의 모델을 도출하였다. 이후 사후분석을 통해, 국내의 가장 대표적인 O2O(Online to Offline) 플랫폼인 네이버 쇼핑의 2024년 실제 데이터에 RFM Framework 기반의 K-Means 클러스터링을 통해 분석하였다. 결과적으로 본 연구는 고객군을 각 감정 점수를 반영하여 세분화함으로써 CRM(Customer Relationship Management) 측면에서 비재구매 고객을 관리하기 위해서 극단적인 부정적 감정에 해당하는 분노, 혐오가 발생하지 않도록 VoC(Voice of Customer)에 대한 처리나, 제품 구매후 서비스(A/S)를 통해 관리하는 것이 중요하다는 점을 제시하였다. The development of online platforms in recent years has catalyzed significant growth in digital commerce. Consequently, the importance of online purchase reviews has become prominent for businesses and consumers alike. However, previous research in Customer Sentiment Analysis (CSA) and Repurchase Prediction has shown limitations, primarily focusing on binary sentiment analysis or utilizing only some textual or structured data. To address these limitations, this study undertakes two phases. In Phase 1, a 7 sentiment classification model is trained to classify seven emotions—Anger, Disgust, Fear, Happiness, Sadness, Surprise and Neutrality—with an accuracy of 93%. Subsequently, in Phase 2, a repurchase prediction model is developed reflecting the seven emotions derived from Phase 1. This model achieves a notable accuracy of 90.25%, outperforming the model that does not incorporate emotions with an accuracy of 89.04%. Following this, in the subsequent post-analysis, RFM Framework-based K-Means clustering is employed on real data from Naver Shopping, a prominent O2O platform in South Korea, from the year 2024. In conclusion, this study highlights the critical role of managing extreme negative emotions such as Anger and Disgust in CRM (Customer Relationship Management) to effectively address non-repurchasing customers. Strategies can include proactive handling of customer dissatisfaction through Voice of Customer (VoC) initiatives and enhancing post-purchase services to optimize product quality.

An Analysis on the Consumption Structure of Contemporary Popular Culture with Network Science

이정우 포항공과대학교 융합대학원 2022 국내석사

Globalization and the development of information technology have enabled people in different societies to share their culture and consume cultural products through digital devices. This social change has made contemporary popular culture to transcend the borders between countries and penetrate the daily lives of consumers. Our thesis focused on investigating which social factors affect the consumption structure of contemporary popular culture. We constructed a consumption network of mobile games between countries to reflect the characteristics of contemporary popular culture. Using Hofstede's cultural dimensions theory and Facebook's social connectedness index, we revealed cultural distance and social ties between countries play important roles in shaping the consumption structure.

Motion planning with planar geometric models

문상룡 포항공과대학 1991 국내석사

PCA?HDBSCAN Fusion Method for Enhanced Dense Target Discrimination in Satellite SAR Imagery

김도수 포항공과대학교 융합대학원 2026 국내석사

Dense-target discrimination in high-resolution synthetic aperture radar (SAR) imagery remains a difficult problem because speckle noise, spatially varying back- ground statistics, and the overlap of closely spaced scatterers reduce target sep- arability. These effects distort local structures and increase sensitivity to al- gorithmic parameters. Classical density-based methods such as DBSCAN and PCA–DBSCAN show abrupt changes in cluster geometry with small parameter variations, often merging multiple dense targets and producing unstable recall and F1-score across scenes. This thesis presents a PCA–HDBSCAN framework that improves structural normalization and clustering stability. PCA alignment linearizes dense-target arrangements and mitigates orientation-related distortions, creating a more con- sistent representation prior to clustering. Hierarchical density estimation then as- signs stability-based persistence scores that preserve true target multiplicity while suppressing spurious detections. This approach yields reliable clusters across a broader range of density thresholds and parameter settings. A multi-level evaluation protocol is also developed, integrating pixel-level ROC analysis, cluster-level stability assessment, and object-level multiplicity es- timation. ROC metrics highlight performance in low–false-alarm regimes, stabil- ity measures quantify cluster persistence across scales, and multiplicity metrics evaluate the prevention of under-segmentation in densely populated regions. To- gether, these analyses provide a unified and interpretable framework for assessing dense-target discrimination performance. Experiments on diverse SAR scenes demonstrate consistent improvements in recall, cluster stability, and F1-score compared with DBSCAN and PCA–DBSCAN. Overall, the combination of PCA-based structural normalization and stability- driven hierarchical clustering provides a robust and interpretable basis for dense- target discrimination in operational SAR environments.

A Study on Personalized Blood Glucose Forecasting and Segmentation using Domain-Specific Variables

이소민 포항공과대학교 융합대학원 2024 국내석사

This study focuses on the importance of short-term glucose prediction in diabetes management facilitated by Continuous Glucose Monitoring (CGM). It recognizes the pivotal role of variables such as diet and insulin intake in glycemic control, while also acknowledging the challenges in effectively collecting this information. To address the gaps presented by missing data, the research prioritizes the identification and application of surrogate markers. These markers indirectly reflect the impact of dietary and physical activity on blood glucose levels, thereby enhancing the accuracy of short-term predictions and contributing to personalized diabetes management. The proposed short-term prediction method incorporates exogenous variables, such as Time in Range (TIR), Continuous Overlapping Net Glycemic Action (CONGA), Mean Amplitude of Glycemic Excursion (MAGE), and Coefficient of Variation (CV). This study presents the framework for glucose prediction and self-management in the context of missing data. This approach enhances the accuracy of short-term predictions and contributes to personalized diabetes management.

A Study on the Improvement of Indicators for Clinical Decision Support Using Electronic Health Record Data

김혜영 포항공과대학교 융합대학원 2024 국내석사

본 연구는 전자의무기록 데이터와 머신러닝 모델을 활용하여 의료진의 임상 의사결정을 보조하는 지표를 개선시키기 위한 방법론을 제시하는 것을 목적으로 한다. 이를 위해, 중환자실에서의 충분한 의료 처치 여부를 기준으로 환자의 중환자실 퇴실 가능 여부를 판단하는 환자 상태 평가 지표의 개선을 사례 연구로 활용하였다. 전자의무기록 데이터와 머신러닝 모델을 활용한 지표 개선 과정은 다음과 같다. 첫째, 개선할 지표의 목적에 부합하는 예측 문제를 설정한다. 둘째, 해당 문제의 결과를 예측하기 위한 머신러닝 모델을 학습시킨다. 셋째, 모델에 활용된 중요 변수를 분석하고, partial dependence를 활용하여 지표 항목을 생성한다. 마지막으로, 모델의 중요 변수들과 기존 임상 현장에서 사용되던 지표의 항목들을 비교 분석하여 지표의 개선점을 탐색한다. 본 연구는 중환자실 퇴실 판단을 위한 환자 상태 평가 지표의 개선을 예시로 위에서 제시된 방법론의 결과를 확인하고자 하였다. 먼저, 중환자실 재입실 시점 간격에 따른 로지스틱 회귀 모델과 오즈비의 비교를 통해, 중환자실 퇴실 후 48시간 이내의 재입실이 중환자실의 의료 서비스가 환자에 미치는 영향을 잘 고려하는 재입실 정의임을 확인하였다. 이어서, 중환자실 퇴실 이후 48시간 이내의 재입실을 예측하는 머신러닝 모델들을 개발하고, 이 모델들의 성능이 베이스라인으로 볼 수 있는 기존의 퇴실 판단 지표보다 우수하게 나타내는 것을 확인하였다. 로지스틱 회귀, 랜덤 포레스트, 나이브 베이즈, 그레디언트 부스팅 등의 머신러닝 모델들 중 그레디언트 부스팅 모델이 가장 높은 예측 성능을 보였다. 마지막으로, 그레디언트 부스팅 모델을 활용하여 중환자실 환자들의 퇴실 판단에 유용한 중요 변수들을 추출하고, partial dependence 분석을 통해 이러한 변수들의 영향을 검토하였다. 이러한 변수들을 기반으로 한 퇴실 판단 지표의 예측은 기존 지표를 통한 예측보다 개선된 성능을 보였다. 또한 기존 퇴실 판단 지표의 변수들과는 달리, 환자 상태 경고 알람 횟수와 같은 새로운 변수들이 중요 변수로 도출되었으며, 이는 지표의 보완에 기여할 수 있음을 확인하였다. 본 연구에서 활용한 임상 의사결정 보조 지표의 개선 방법론은 병원 내의 다양한 임상 의사결정 보조 지표의 개선에 활용되어, 병원 내 의료서비스 관리에 기여할 수 있을 것이다. 복잡한 모델이 끊임없이 개선되고 활용될 수 없는 병원 환경에서도 이러한 방법론을 통한 지표의 개선은 의료진이 보다 정확하고 신속한 의사결정을 내리는 데 도움이 될 것으로 기대된다. This study aims to present a method to improve indicators that support medical staff in clinical decision-making, by using Electronic Health Record (EHR) data and machine-learning (ML) models. Specifically, it focuses on improving patient status assessment indicators that determine the suitability of intensive care unit (ICU) discharge by quantifying the sufficiency of medical treatments provided in the ICU. The process of improving the indicators by using EHR data and ML involves four steps: (1) define a predictive problem that aligns with the purpose of the indicator to be improved; (2) build an ML model to predict outcomes for this defined problem; (3) interpret the important variables of the model and create indicator items by applying partial dependence analysis; (4) identify potential improvements by comparing the model’s significant variables to the existing clinical indicators. As an example of the proposed method, this study focused on improving patient status assessment indicators for ICU discharge decision-making. Initially, a comparison of logistic regression models and odds ratios that consider the interval of ICU readmission, confirmed that readmission within 48 hours after ICU discharge is a valid criterion to evaluate the effectiveness of ICU medical services on patients. Subsequently, ML models were developed to predict readmissions within 48 hours post-ICU discharge; these models were more accurate than the existing discharge-decision indicators, which can be considered as the baseline. Among various ML models, the Gradient Boosting model was had the highest predictive accuracy. Finally, using the Gradient Boosting model, important variables useful for ICU patient discharge decisions were extracted, and their effects were assessed using partial-dependence analysis. The discharge decision indicators that consider these variables showed higher accuracy than the existing indicators. Additionally, new variables (e.g., frequency of patient-status warning alarms) that differed from those in the existing discharge decision indicators, were identified as important; this result confirms their potential contribution to improving the indicators. The clinical decision-support indicator improvement method that is developed in this study can be applied to enhance various clinical decision support indicators within hospitals, contributing to better management of hospital medical services. Even in hospital environments in which complex models cannot be continuously improved and utilized, the improvement of indicators by using this method is expected to aid medical professionals to increase the speed and accuracy and of their decisions.