Earticle

Purpose Climate change affects human health and calls for health meteorological services. The purpose of this study is to find the significant predictors for the demands of the health meteorological information. Methods This study used a descriptive design through structured self-report questionnaires. Data from 956 participants who were at least 18 years old and living in Busan, Korea, were collected from June 1 to July 31, 2009. The data was analyzed using a decision tree method, one of the data mining techniques by SAS 9.1 and Enterprise Miner 4.3 program. Results Two hundred and ninety participants (30.3%) demanded the information, and 505 of them (52.8%) perceived the necessity of health meteorological information. From the decision tree method, the predictors related to the demands of the health meteorological information were determined as “the perception of the necessity of health meteorological information,” “the coping to the weather warnings” and “the importance of the weather forecasting in daily life.” In Particular, the significant different variables in the perception of the necessity of health meteorological information were “female,” “aged over 40” and “environmental diseases.” Thus, the model derived in this study is considered for explaining and predicting the demands of health meteorological information. Conclusions It can be effectively used as a reference model for future studies and is a suggested direction in health meteorological information service and policy development. We suggest health forecasting as a nursing service and a primary health care network for healthier and more comfortable life.

Spiking neural networks (SNN), employing eventbased spike computation, can be implemented in hardware where on-chip learning and inference are supported in a powerand area-efficient manner. Although many SNN hardware have been proposed for energy-efficient designs using relatively shallow networks, SNN algorithms that support multi-layer learning need to be implemented in hardware to handle more complex datasets. However, multi-layer learning requires more complicated functions like softmax activation, which makes energy-efficient hardware design difficult. In this paper, we present a zero-spike prediction method to skip the complicated function in the convolution layer. Decomposing the original algorithm, the proposed method skips at least 76.90% of softmax activation operations without classification accuracy degradation.

주가는 사람들의 심리를 반영하고 있으며, 주식시장 전체에 영향을 미치는 요인으로는 경제성장률, 경제지료, 이자율, 무역수지, 환율, 통화량 등이 있다. 국내 주식시장은 전날 미국 및 주변 국가들의 주가지수에 영향을 많이 받고 있으며 대표적인 주가지수가 다우지수, 나스닥, S&P500이다. 최근 주가뉴스를 이용한 주가분석 연구가 활발히 진행되 고 있으며, 인공지능 기반한 분석을 통하여 과거 시계열 데이터를 기반으로 미래를 예측하는 연구가 진행 중에 있다. 하지만, 주식시장은 예측시스템에 의해서 단기간 적중이 되더라도, 시장은 더 이상의 단기 전략대로 움직여지지 않고, 새롭게 변할 수밖에 없다. 따라서, 본 모델을 삼성전자 주식데이터와 뉴스 정보를 텍스트 마이닝으로 모니터링하여 분 석한 결과를 나타내어 예측이 가능한 모델을 제시하였으며, 향후 종목별 예측을 통하여 실제 예측이 정확한지 확인하여 발전시켜 나갈 예정임

The stock price reflects people's psychology, and factors affecting the entire stock market include economic growth rate, economic rate, interest rate, trade balance, exchange rate, and currency. The domestic stock market is heavily influenced by the stock index of the United States and neighboring countries on the previous day, and the representative stock indexes are the Dow index, NASDAQ, and S & P500. Recently, research on stock price analysis using stock news has been actively conducted, and research is underway to predict the future based on past time series data through artificial intelligence-based analysis. However, even if the stock market is hit for a short period of time by the forecasting system, the market will no longer move according to the short-term strategy, and it will have to change anew. Therefore, this model monitored Samsung Electronics' stock data and news information through text mining, and presented a predictable model by showing the analyzed results.

In the dynamic world of e-commerce, effectively deciphering customer reviews is of paramount importance. This study uniquely combines the RFE-SHAP feature selection technique with topic modeling (LDA) to address prevalent challenges like overfitting in predictive modeling. Our empirical analysis underscores the superior performance of the Random Forest model, particularly when refined with a subset of 14 pivotal features. Notably, topics such as quality and appearance, fit and comfort, and durability concerns emerged as significant determinants of customer satisfaction within the clothing sector. Utilizing data exclusively from Amazon's clothing reviews, our research emphasizes the criticality of strategic feature selection and delves deep into the multifaceted factors shaping customer sentiments. By seamlessly merging quantitative metrics with qualitative content insights, this study not only offers a robust framework for understanding online reviews but also paves the way for future research in optimizing e-commerce strategies based on customer feedback.

Structure-based drug design possibly benefit from in silico methods that precisely predict the binding affinity of small molecules to target macromolecules. There are many limitations arise from the difficulty of predicting the binding affinity of a small molecule to a biological target with the current scoring functions. There is thus a strong interest in novel methodologies based on MD simulations that claim predictions of greater accuracy than current scoring functions, helpful for a regular use designed for drug discovery in the pharmaceutical industry. Herein, we report a short review on free energy calculations using MMPBSA method a useful method in structure based drug discovery.

최근 LSTM(Long Short-Term Memory)모델에 대한 연구는 강수-유출 분석에서의 성능은 기존 물리 기반 강우-유출 모델에 비해 더 우수한 성과를 보여주고 있습니다.. 그러나 데이터 품질관리는 LSTM과 같은 Data-Driven모델에서는 모델의 성능 향상을 위해 여전히 중요한 요소입니다. 본 연구에서는 기존 단순 물수지 분석법(Simple Water Balance method)으로 추정된 시간단위 유입량의 변동성을 감소시키기 위해 이벤트 식별(Event Identification) 및 지수 평활화(Exponential Smoothing) 방법(EI&ES)을 적용하여 데이터 전처리를 하였습니다. 이후 (1) 기존방법, (2) 이동평균, (3) EI&ES방법으로 12가지 Cases에 대해서 LSTM모델을 NSE(Nash-Sutcliffe Efficiency), RMSE(Root Mean Square Error), MSE(Mean Squared Error)와 같은 성능 지표로 평가한 결과, EI&ES방법이 높은 성능을 보임을 알 수 있었습니다. 본 연구는 새로운 딥러닝 방법뿐만 아니라, 딥러닝 성능을 향상시키는 데 중요한 데이터 중심 연구를 강수-유출 분석의 사례로 제시하고 있습니다.

The performance of LSTMs in analyzing rainfall-runoff dynamics has demonstrated better performance over the physically-based hydrological models. However, data quality remains crucial to further improve the performance of LSTM. In this study, we applied the Integration of Event Identification & Exponential Smoothing (EI & ES) methods to reduce the fluctuations in inflow estimation, which are typically calculated by the Simple Water Balance method. The preprocessed inflow data was then used to train a LSTM, demonstrating a noticeable improvement with 12 different inflow preprocessing cases in performance metrics such as NSE, RMSE, and MSE for hourly dam inflow prediction. This research presents not only new deep learning methods but also data-centric research that is critical to improving the performance of deep learning as an example of rainfall-runoff analysis.

In autonomous driving, environmental perception and decision-making are technologies that acquire signals based on various sensors and generate information that enables obstacle avoidance, emergency stops, and path planning. Such environmental perception technologies are limited in that they depend on expensive sensors such as LIDAR and RADAR, so research on environmental perception technologies using only cameras is actively being conducted. Studies predicting traffic accidents based on dashcam footage have also been performed as part of such research. This is challenging because only limited forward-looking footage can be acquired, and the surrounding environment is dynamic and changes quickly, making analysis difficult. Existing research has focused on learning the spatialtemporal feature representation to solve these problems. This paper extracts the trajectories of the ego-vehicle and surrounding vehicles using Visual SLAM and Multiple Object Tracking algorithms and uses them as inputs to the graph convolutional neural network(GCN) to learn the spatialtemporal feature representation. In addition, the features learned through the GCN are used as inputs to a bayesian neural network(BNN) to predict the probability of accidents, and its ability to predict accidents in advance has been verified by comparison with existing studies.

본 보고에서는 강우에 의한 대규모 노천광산 사면의 붕괴 위험도 평가에 활용 가능한 기술로서 침투-응력 해석에 의한 사면안정 평가기술 뿐만 아니라 이를 보완하기 위한 방안으로 강우량 자료에 근거한 위험도 추정, 사면에서의 계측자료를 이용한 위험도 예측 기법, 수치표고모델(Digital Elevation Model, DEM)을 이용한 위험도 예측 기법 등을 소개하였다. 누적강우량과 강우강도를 이용하여 사면 붕괴 위험도를 예측할 수 있음을 보였으며, 지하수면 상부 모관흡수력의 계측을 통한 표층 붕괴 위험도 예측이 가능함을 알 수 있었다. DEM을 이용한 사면 붕괴 위험도 예측기법은 대규모 노천 사면의 위험지역을 사전에 추출하여 관리하고 채광작업 진행에 따른 위험도 변화 양상을 신속하게 파악하는데 유용할 것으로 판단된다. 또한, 노천광산 사면의 안정성을 향상시키기 위한 보조공법으로 자연배수, 수평배수공, 양수공법, 배수 터널공과 같은 지하수위 하강공법의 종류 및 특징과 사면의 지질학적 및 수리지질학적 특징에 따른 적용성에 대해 기술하였다.

In this study, we have introduced a risk prediction method for the slope failure at large open pit mine. The method includes the risk prediction based on rainfall data, using monitored data, and in terms of digital elevation model (DEM) in addition to a conventional slope stability analysis as a result of a seepage-stress analysis. Cumulative precipitation rate and intensity were useful in predicting slope failure induced by rainfall and monitoring suction pressure above groundwater table was also applicable. Slope failure prediction due to rainfall using DEM was considered to help identify potentially dangerous regions and their evolution during excavation. And mine dewatering technologies such as natural and horizontal drainage, vertical pumping and drainage tunnel method to enhance a open pit mine slope stability were introduced as well together with their applicability, depending on geological and hydrogeological characteristics of the slope.

지하철, 철도, 도로터널 등에서 터널주변의 지반을 보강하고 막장을 안정시키기 위하여 UAM (Umbrella Arch Method)이 많이 적용되고 있다. 그러나 이 공법은 현장 시공사례를 통한 경험적인 방법에 의 해 설계와 시공이 이루어지고 있어 체계적인 설계·시공방법과 보강효과를 정량적으로 판단할 수 있는 방안 이 요구된다. 본 연구에서는 유한요소법을 이용하여 지반의 탄성계수와 지층두께에 대한 매개변수 연구를 수 행하였으며, UAM의 보강효과를 변위에 대해 정량적으로 분석하였다. 그 결과, UAM 적용시 토사지반의 터 널은 12%~32%, 풍화암지반의 터널은 12%~26% 그리고 연암지반의 터널은 11%~18%의 변위가 감소하였다. 또한, 지반의 탄성계수와 지층두께에 대해 회귀분석을 실시하여 막장면 천단침하량(So)과 수렴후 최대침하량 (Smax)으로 표현되는 지수함수 형태의 천단침하량 예측식을 제안하였다.

Recently, Umbrella Arch Method(UAM) is commonly used in tunnel design in order to reinforce the ground around tunnel and stabilize the tunnel face. But quantitative estimation of reinforcement effect is needed because UAM is designed and constructed only on the basis of empirical experience. By using 3-dimensional finite element method, parametric study is performed for elastic modulus of ground and overburden, and reinforcement effect is analyzed quantitatively. From the results, crown settlement decreases about 12%~32% in soil tunnel, about 12%~26% in weathered rock tunnel and 11%~18% in soft rock tunnel when applied with UAM. The prediction equation for final crown settlement is suggested through regression analysis and the equation is expressed as exponential function which has variables such as So and Smax.

클라우드 컴퓨팅은 IT 인프라의 초기투자비용과 유지보수에 대한 부담을 줄이는 방안으로 적합하다고 여겨지면서발전하였다. 클라우드 컴퓨팅이 발전함에 따라, 그 환경을 유지하기 위해 소모되는 전력에 관한 문제가 발생하였다. 따라서 소비전력을 줄이거나 전력효율을 향상시키기 위한 연구가 진행되고 있다. 하지만 전력 효율을 향상시키기 위해서는 소비전력을 측정해야한다. 하지만 소비전력을 측정하기 위해 전력측정기를 사용하는 것은 추가적인 비용이발생할 뿐만 아니라 전력측정기를 설치해야 하는 번거로움이 존재한다. 따라서 본 논문에서는 전력측정기를 이용하지 않고 소비전력을 예측하는 수식을 제안한다. CPU, Memory, Hard disk의 TDP와 사용률을 이용하여 전력을예측하는 예측 방안을 제안한다. 또한 제안한 수식의 정확성을 입증하기 위하여, CPU 테스트를 수행하여 예측소비전력과 실제소비전력을 비교분석 하였다. 그 결과 수식을 이용한 예측전력과 전력측정기를 이용한 실제전력의 오차율평균은 CPU 테스트에서 5.75%를 보였고, Hard disk 테스트에서는 8.54%를 보였다.

Cloud computing is suited to reduce costs of maintaining IT infrastructure and initial investment, so it has been developed. As cloud computing is developed, problems of power consumption to maintain the environment are occurred. So, several studies for reducing power consumption or improving power efficiency have been performed. However, for improving power efficiency, we have to measure power consumption. For measuring the power consumption by using the power meter has not only additional cost, but also inconvenience for installing power meter. Thus, we propose prediction method of power consumption using utilization rate of CPU, Memory, and Hard disk. Also, to prove accuracy of prediction method, we perform the CPU test to compare and analyze between predictive power consumption and actual power consumption. As a result, there are differences between the predictive power consumption and actual power consumption value about 5.87% in the test.

Social network has been expected to increase the value of social capital through online user interactions which remove geographical boundary. However, online users in social networks face challenges of assessing whether the anonymous user and his/her providing information are reliable or not because of limited experiences with a small number of users. Therefore, it is vital to provide a successful trust model which builds and maintains a web of trust. This study aims to propose a prediction method for the interpersonal trust which measures the level of trust about information provider in Facebook. To develop the prediction method, we first investigated behavioral research for trust in social science and extracted 5 antecedents of trust : lenience, ability, steadiness, intimacy, and similarity. Then we measured the antecedents from the history of interactive behavior and built prediction models using the two decision trees and a computational model. We also applied the proposed method to predict interpersonal trust between Facebook users and evaluated the prediction accuracy. The predicted trust metric has dynamic feature which can be adjusted over time according to the interaction between two users.

본 논문에서는 풍력 예측의 정확도를 높이기 위해 풍향 특성을 고려한 풍력 예측 방법을 제안한다. 제안 된 방법은풍력 특성을 추출하고 발전을 예측하는 것을 포함한다. 바람의 특성을 추출하기 위해 발전량, 풍향 및 풍속의 상관분석이 수행된다. 풍향과 풍속의 상관 관계를 바탕으로 K-means 법을 이용한 클러스터링을 통해 특징 벡터를 추출한다. 예측 부분에서는 임의의 실수 값을 예측할 수 있도록 SVM (Support Vector Machine)을 일반화 한 SVR (Support Vector Regression)을 사용하여 기계 학습을 수행한다. 제안 된 방법의 정확성과 타당성을 검증하기위해 제주도 풍력 발전 단지의 3 곳에서 수집 한 자료를 이용하였다. 실험 결과는 제안 된 방법의 오차가 기존 풍력발전 방법의 오차보다 우수하다는 것을 보여준다.

In this paper, we propose a wind power forecasting method that takes into consideration wind characteristics to improve the accuracy of wind power prediction. The proposed method involves extracting wind characteristics and predicting power generation. Correlation analysis of power generation amount, wind direction, and wind speed is performed for extracting wind characteristics. Based on the correlation between the wind direction and the wind speed, the feature vector is extracted by clustering using the K-means method. In the prediction part, machine learning is performed using the SVR (support vector regression) that generalizes the SVM (support vector machine) so that an arbitrary real value can be predicted. To verify the accuracy and feasibility of the proposed method, we used the data collected from three different locations of the Jeju Island wind farm. Experimental results show that the error of the proposed method is better than that of existing wind power generation methods.

Recently, developing and applying of the double skin, which can utilize the natural energy with more aggressive characteristic, is now in study. The studies of the double skin are been constantly in progress in advanced countries as well as in domestic. And the double skin has lately been put to practical use. This study has selected the building with the box-typed double skin system and performed the thermal performance prediction of double skin system using TRNSYS. The U-value of the double skin system has been calculated. And this one has compared the measured U-value with the simulated U-value of double skin system. The results of comparison between experiment and the simulation were similar. So a prediction method on thermal performance of the box-typed double skin using TRNSYS simulation has secured a feasibility.

본 논문은 천리안 위성을 이용하여 운용중인 공공 테스트베드의 가용율 및 시스템 throughput 향상을 위해 포워드 링크 ACM(Adaptive Coding & Modulation) 방안에 대해 소개하고, 포워드링크에 ACM 기능을 구현하기 위한 채널 상태를 예측하기위한 알고리즘으로 기울기를 이용한 예측 기법과 LMS(Least Mean Square)를 이용한 예측 기법의 성능을 비교하였다. 시뮬레이션 결과 LMS 기법을 이용한 예측기법은 99%가 3dB 이내의 예측 오차를 보였고, 기울기를 이용한 예측 기법은 99%가 4.5dB 이내의 예측 오차를 갖음을 알 수 있다.

In this paper, we present the forward link ACM method to improve the link availability and system throughput. Also, we compare the prediction algorithm between slope based prediction and LMS algorithm. The simulation results show that the 99% of predicted values in LMS algorithm is within 3dB and that of predicted values in the slope based prediction method is within 4.5dB.

최근 자율주행, 로봇 등에 장착된 카메라 관련 이미지 처리 모델이 각광받고 있다. 카메라에서 얻을 수 있는 연속된 이미지의 다음 장면을 예측하기 위해 생성 모델과 RNN 계열 모델을 결합한 모델 연구가 활발하다. 그 중 이미지 픽셀 학습 모델이 우수한 성능을 보이나, 복잡한 구조로 인해 개선 부분을 명확 히 설명하기 어렵고 계산 비용이 증가한다. 본 논문은 얽힘 없는 잠재 변수의 학습을 통해 모델의 해석 가능성을 향상시키며, 동시에 단순화된 구조를 채택하는 예측 기법을 제시한다