Earticle

Prediction of blood glucose (BG) values in type 1 diabetes (T1D) remains an essential and challenging issue. Recently, machine learning methods have been used to solve this problem. We present a forecasting model based on extreme gradient boosting (XGBoost) regression to estimate the BG value of T1D patients in this study. We developed the models using clinical datasets from five T1D patients and forecasted the BG value for the following prediction horizons (PHs) of 10, and 20 minutes. Datasets are divided in two parts, around 60:40 for training and testing. We compared the performance of our proposed model to existing models using several performance metrics, including root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R2). Our suggested XGBoost model outperformed existing models, with average RMSE, MAPE, and R2 of 12.57 mg/dL, 7.15%, 0.94, and 21.93 mg/dL, 13.72%, 0.84 for PH of 10 and 20 minutes, respectively. Finally, the findings of this study are intended to be applied to improve diabetes care.

We need a study of the ratio of successful applicants control methodology about the national technical qualification under the global-green industrial society and rapid change of international circumstances, infinite competition rider society under FTA aspects. It is necessary to develop of HRD Korea selfishness and increase brand value of national technical qualification. So, this study is analysed to the ratio of successful applicants of national technical qualification toward change of the 'bank of problems' control rule, various characters of candidates and the trend of demand and supply of labours instead of the absolute evaluation method. Accordingly, this study suggests to a methodology for the forecasting model of the ratio of successful applicants using the level of problems difficulty and pattern and the candidates academical carriers.

We develop a model to predict traffic accidents in Korea. In contrast to the classical approach that mainly uses regression analysis, Bayesian approach is adopted. A dependent model that incorporates the data from different kinds of accidents is introduced. The rate of severe accident can be updated even with no data of the same kind. The data of minor accident that can be obtained frequently is efficiently used to predict the severe accident.

This study introduces an advanced artificial model, grey wolf optimization (GWO)-nested complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN)-convolutional neural network (CNN)-bidirectional long short-term memory (BiLSTM), for wind speed forecasting. Initially, CEEMDAN with two nested layers decomposes the time series into intrinsic mode functions (IMFs) to enhance forecasting capabilities. Subsequently, CNN extracts features from IMFs, and BiLSTM captures temporal dependencies for precise predictions. GWO further enhances the accurac by selecting optimal hyperparameters based on decomposition results. Test results on diverse wind speed datasets demonstrate the model’s superiority, with a mean absolute percentage error (MAPE) of approximately 3%.

This study addresses default prediction through ensemble techniques and cost-sensitive learning. It introduces a novel stacking method, focusing on instances with varying misclassification costs. Previous research lacks this comprehensive approach, highlighting a research gap. The proposed technique proves advantageous in terms of economic payoff and performance, offering practical utility. It complies with regulatory monitoring standards without incurring statistical cost penalties. This study demonstrates potential revenue gains even without precise cost ratios. Empirical results using Taiwanese company bankruptcy data (1999-2009, 95 financial ratios) reveal significant outcomes. Firstly, the proposed algorithm markedly improves economic payoff. Secondly, its statistical performance remains unaffected, even considering the dependency on overall prediction errors related to misclassifying defaults. Lastly, the method is computationally efficient, robust across cost scenarios.

최근 다양한 국가에서 영향예보를 위한 연구가 수행되고 있다. 이러한 영향예보는 정확한 기상예보 를 기반으로 다양한 분야의 침수피해를 예방하기 위해 활용이 가능하다. 본 연구에서는 부산광역시 를 대상으로 하천의 호우영향예보 모형을 개발하였다. 호우영향예보 모형은 1km x 1km의 격자를 대상으로 하천의 홍수위와 한계강우량에 대하여 산정식을 개발하였다. 한계강우량은 하천의 종점을 기준으로 설계홍수량과 설계홍수위에 대한 유량-수위 관계를 3시간 지속시간의 한계강우량에 수위- 강우에 대하여 산정식을 분석하였다. 모형의 검증은 2020년 7월 22일부터 24일까지 총 283.3mm가 발생한 호우사상을 분석하였다. 분석결과 하천의 호우영향예보는 7월 23일 22시부터 23시까지 호우 영향레벨이 계획홍수량의 60%~70%(MEDIUM)이 발령되었다. 호우영향예보가 발령된 하천에서는 도로침수, 고수부지 침수 등의 피해자료와 일치하여 모형의 적정성을 확인하였다. 또한, 수영강 수위 관측소의 홍수위와 호우영향예보의 발령기준이 동일한 것으로 검토되어 모형의 정확도가 높은 것으로 검증되었다.

Impact forecasting has been applied to estimate flood damage in many different countries based on accurate weather forecasts. This study develops a model forecasting the impact of rainfall on riverside flood risks in the Busan Metropolitan City. The rainfall impact forecasting equations were obtained on a 1km x 1km grid with respect to water levels and critical rainfall thresholds. Critical rainfall was determined based on the end point of river and the discharge-water level relationship between design flood discharge and design flood level by analyzing water-rainfall in 3-hour critical rainfall. Model verification was performed using a heavy rain event with a total of 283.3 mm from July 22 to 24, 2020. According to the rainfall impact forecasting, 60 to 70% of the design flood level (“MEDIUM” alert) was issued on rivers between 10-11pm on July 23. In the rivers where the rainfall impact alert was issued, actual damages were found such as flooded roads and floodplain, which confirmed the appropriateness of the model. The high model accuracy was also validated by comparing the alert standard for water level with forecasted rainfall impact at the water level station in Suyeong River.

본 연구에서는 수도권 4개 권역의 PM2.5 예보를 위해 심층신경망 모델을 개발하고, 계층별 관련성 전파를 사용하여 예 측결과에 대한 입력인자들의 기여도를 분석했다. 제안한 심층신경망 예보모델은 평균적으로 지수적중률 71%, 고농도 감 지 확률 65%, 오경보율 40% 정도의 성능을 보였다. 이러한 심층신경망 예보결과에 대한 입력인자들의 중요도를 분석하 기 위해 데이터를 지역 및 PM2.5 농도로 분류하여 계층별 관련성 전파를 수행했다. 그 결과 농도 및 지역에 상관없이 중 요도가 높은 인자는 압력, 온도, 이슬점온도, 상대습도, U, V, 이산화질소, 일산화탄소로 확인된 반면에 누적강수량, 아황 산가스, PM10, 오존은 중요도가 낮은 것으로 관찰되었다. 고농도 사례에서는 일사와 오존의 중요도가 낮아지고, 예측시 간에 근접한 시간대의 PM2.5의 중요도가 높아졌으며, CMAQ(The Community Multiscale Air Quality) 예측인자들의 중 요도도 다소 올라갔다. 지역별 입력인자들의 중요도는 대체로 유사했는데 인천과 경기북부의 경우 일사의 중요도가 낮아 지는 등 고농도 패턴의 인자중요도와 일부 비슷한 결과를 보였다. 이것은 인천과 경기북부가 다른 지역보다 고농도 데이 터가 더 자주 발생하기 때문으로 분석된다. 이러한 인자중요도 결과는 향후 고농도 적중률 향상 및 지역별 특성에 적합 한 예보모델을 개발하는데 활용하면 효과적일 것으로 기대한다.

In this study, we propose a deep neural network model for PM2.5 prediction in the four metropolitan areas, and analyze the contribution of input factors to the prediction results using layer-wise relevance propagation. The proposed deep neural network forecasting model exhibited an average exponential accuracy of 71%, a high-concentration detection probability of 65%, and a false alarm ratio of 40%. In order to analyze the importance of input factors for these deep neural network forecasting results, we classify the data into regional and PM2.5 concentration level and performed layer-wise relevance propagation. As a result, factors such as pressure, temperature, dew point temperature, relative humidity, U, V, nitrogen dioxide and carbon monoxide were found to be important, regardless of any region and concentration level. On the other hand, accumulated precipitation, sulfur dioxide, PM10, and ozone are shown to have lower importance. In highconcentration cases, the importance of solar radiation and ozone decreased, PM2.5 during time periods close to the prediction time increased, and the importance of CMAQ(The Community Multiscale Air Quality) predictive factors also showed some degree of elevation. The importance of input factors by region are generally similar; however, in the case of Incheon and Northern Gyeonggi Province, the importance of the radiation decreases, showing some similarities to the importance of factors in high concentration patterns. This is analyzed as being due to a higher occurrence of highconcentration data in Incheon and Northern Gyeonggi Province compared to other regions. These analyzing results are expected to be effective in developing forecasting models that improve the prediction accuracy of high concentration and match regional characteristics in the future.

In this study, Mike11 will be used as the numerical model where a data assimilation method will be applied to it. This paper aims to gain an insight and understanding of data assimilation in flood forecasting models. It will start with a general discussion of data assimilation, followed by a description of the methodology and discussion of the statistical error forecast model used, which in this case is the linear regression. This error forecast model is applied to the water level forecast simulated by MIKE11 to produced improved forecast and validated against real measurements. It is found that there exists a phase error in the improved forecasts. Hence, 2 general formula are used to account for this phase error and they have shown improvement to the accuracy of the forecasts, where one improved the immediate forecast of up to 5 hours while the other improved the estimation of the peak discharge.

The implied volatility from Black and Scholes (1973) model has been empirically tested for the forecasting performance of future volatility and commonly shown to be biased. Based on the belief that the implied volatility from option prices is the best estimate of future volatility, this study tries to find out a better model, which can derive the implied volatility from option prices, to overcome the forecasting bias from Black and Scholes (1973) model. Heston (1993)’s model which improves on the problems of Black and Scholes (1973) model the most for pricing and hedging options is one candidate, and VIX which is the expected risk neutral value of realized volatility under the discrete version is the other. This study conducts a comparative analysis on the implied volatility from Black and Scholes (1973) model, that from Heston (1993)’s model, and VIX for the forecasting performance of future volatility. From the empirical analysis on KOSPI200 option market, it is found that Heston (1993)’s implied volatility eliminates the bias mostly which Black and Scholes (1973) implied volatility has. VIX, on the other hand, does not show any improvement for the forecasting performance.

최근 들어 IITS는 스마트 시티관련 산업계에서 중요한 주제로 떠오르고 있다. IITS의 주요 목적인 교 통체증 (차량 사고에 기인한) 예방책들이 발전된 센서 및 통신 기술의 도움을 받아 다양하게 시도되었다. 관 련 연구들에서는 자동차 사고와 사고 위치적 특성, 날씨, 운전자 행동, 시간 등 다양한 요인들과 상관 관계가 있음을 보여주고 있다. 우리 연구는 자동차 사고와 사고 발생 시간 사이의 상관관계에 주제를 집중했다. 본 논문에서는 ARIMA (Auto-Regressive Integrated Moving Average) 자동 회귀, 정상 및 지연 순서를 결정 하는 세 가지 요소를 확인하기 위해 ADF (Augmented Dickey-Fuller)를 포함한 ARIMA 테스트를 수행했 다. 본 연구 결과로서 시간 별 자동차 충돌 수 예측에 대한 요약을 제시하며, 한국 내 자동차 사고 데이터는 ARIMA 모델에 적용될 수 있음을 보여주었고, 국내 자동차 사고는 하루를 기준으로 일정한 주기가 존재하는 성격을 가지고 있다는 것을 제시했다.

Recently, IITS (intelligent integrated transportation system) has been important topic in Smart City related industry. As a main objective of IITS, prevention of traffic jam (due to car accidents) has been attempted with help of advanced sensor and communication technologies. Studies show that car accident has certain correlation with some factors including characteristics of location, weather, driver’s behavior, and time of day. We concentrate our study on observing auto correlativity of car accidents in terms of time of day. In this paper, we performed the ARIMA tests including ADF (augmented Dickey-Fuller) to check the three factors determining auto-regressive, stationarity, and lag order. Summary on forecasting of hourly car crash counts is presented, we show that the traffic accident data obtained in Korea can be applied to ARIMA model and present a result that traffic accidents in Korea have property of being recurrent daily basis.

본 연구에서는 earth dam의 파괴로부터 야기되는 유출수문곡선을 유도하고, 홍수파가 자연하도를 통하여 하류로 전파되어 나가는 과정을 추적하기 위하여 dynamic wave equation을 구성하고, 이 식을 Preissmann scheme과 double sweep법을 이용하여 해석하는 DBF(Dam Break Floodwave) 요형을 개발하였다. 본 모형을 Teton dam에 적용하여 댐 파괴로 인한 홍수파의 첨두수위, 청두유량, 도달시간 및 범람 범위 등을 계산하였고, 실측자료와의 비교검토를 통하여 본 모형의 적용성을 입증하였다.

The floodwave propagation resulting from the earth dam-break is studied. DBF(Dam-Break Floodwave) model based on the dynamic wave equation is presented by introducing Preissmann scheme and double sweep algorithm. DBF model is applied to the Teton dam, and the numerical results have good agreements with the data observed in the peak elevation profile, the peak discharge, the flood travel time and the flooded area.

The series of the papers consist of three parts to describe the development, calibration, and applications of the flood forecasting models for the Youngsan Estuarine Dam located at the mouth of the Youngsan river. And this paper discusses the hydrologic model for inflow simulation at Naju station, which constitutes 64 percent of the drainage basin of 3521 .6km$^2$ in area. A simplified TANK model was formulated to simulate hourly runoff from rainfall And the model parameters were optirnized using historical storm data, and validated with the records. The results of this paper were summarized as follows. 1. The simplified TANK model was formulated to conceptualize the hourly rainfall-run-off relationships at a watershed with four tanks in series having five runoff outlets. The runoff from each outlet was assumed to be proportional to the storage exceeding a threshold value. And each tank was linked with a drainage hole from the upper one. 2. Fifteen storm events from four year records from 1984 to 1987 were selected for this study. They varied from 81 to 289rn'm The watershed averaged, hourly rainfall data were determined from those at fifteen raingaging stations using a Thiessen method. Some missing and unrealistic records at a few stations were estimated or replaced with the values determined using a reciprocal distance square method from abjacent ones. 3. An univariate scheme was adopted to calibrate the model parameters using historical records. Some of the calibrated parameters were statistically related to antecedent precipitation. And the model simulated the streamflow close to the observed, with the mean coefficient of determination of 0.94 for all storm events. 4. The simulated streamflow were in good agreement with the historical records for ungaged condition simulation runs. The mean coefficient of determination for the runs was 0.93, nearly the same as calibration runs. This may indicates that the model performs very well in flood forecasting situations for the watershed.

기상 자료 미계측 지점의 단기 강우 예보 모형을 개발하였다. 본 연구 모형은 강우 모의 모형, 기상학적 동질성, 그리고 기상 변수 예측 및 추정에 관한 몇 가지 가정을 전제로 하였으며 강우의 예보에는 칼만 필터 기법을 사용하였다. 기존 모형의 방정식은 수운적 크기 분포(HSD)가 강우 강도에 종속이므로 강우량에 대하여 비선형이다. 본 연구 모형의 방정식은 HSD를 구름층 저류량의 함수로 구성함으로써 강우량에 대하여 비선형이다. 본 연구 모형의 방정식은 HSD를 구름층 저류량의 함수로 구성함으로써 강우량에 대하여 선형화되었다. 또한 기상 입력 변수는 경험 모형에 의하여 예측되었다. 본 연구 모형을 대청댐 유형의 호우 사상에 적용하였다. 그 결과 예보 및 실측 강우 강도간의 평균 자승 오차는 0.30~1.01 mm/hr이었다. 이 결과로 미루어 볼 때, 본 연구 모형에 수반된 가정은 합리적이며 본 연구 모형은 기상 자료 미계측 지점에서 강우를 단기 예보하는데 유용하다고 판단된다.

The rainfall forecasting model of the short term is improved at the point where meterological data is not gaged. In this study, the adopted model is based on the assumptions for simulation model of rainfall process, meteorological homogeneousness, prediction and estimation of meteorological data. A Kalman Filter technique is used for rainfall forecasting. In the existing models, the equation of the model is non-linear type with regard to rainfall rate, because hydrometer size distribution (HSD) depends on rainfall intensity. The equation is linearized about rainfall rate as HSD is formulated by the function of the water storage in the cloud. And meteorological input variables are predicted by emprical model. It is applied to the storm events over Taech'ong Dam area. The results show that root mean square error between the forecasted and the observed rainfall intensity is varing from 0.3 to 1.01 mm/hr. It is suggested that the assumptions of this study be reasonable and our model is useful for the short term rainfall forecasting at the ungaged point of the meteorological data.

The basic concept of the model is to minimize the error range between forecasted flood inflow and actual flood inflow, and forecast accurately the flood discharge some hours in advance depending on the concentration time(Tc) and soil moisture retention storage(Sa). Simplex method that is a multi-level optimization technique was used to search for the determination of the best parameters of RETFLO (REal-Time FLOod forecasting) model. The flood forecasting model developed was applied to several strom event of Yedang reservoir during past 10 years. Model perfomance was very good with relative errors of 10% for comparison of total runoff volume and with one hour delayed peak time.

One of the most difficult problem to estimate the flood inflow is how to understand the effective rainfall. The effective rainfall is absolutely influenced by the condition of soil moisture in the watershed just before the storm event. DAWAST model developed to simulate the daily streamflow considering the meteologic and geographic characteristics in the Korean watersheds was applied to understand the soil moisture and estimate the effective rainfall rather accurately through the daily water balance in the watershed. From this soil moisture and effective rainfall, concentration time, dimensionless hydrograph, and addition of baseflow, the rainfall-runoff model for flood flow was developed by converting the concept of long-term runoff into short-term runoff. And, real-time flood forecasting model was also developed to forecast the flood-inflow hydrograph to the river and reservoir, and called RETFLO model. According to the model verification, RETFLO model can be practically applied to the medium and small river and reservoir to forecast the flood hydrograph with peak discharge, peak time, and volume. Consequently, flood forecasting and warning system in the river and the reservoir can be greatly improved by using personal computer.

홍수위 예측의 정확도를 향상시키기 위하여 분포형 매개변수를 갖는 동수역학적 흐름모형과 심화학습 기반의 오차보정 모형을 결합한 혼합모형을 제시하였다. 한강 하류구간을 대상으로 동수역학적 계산모형을 수립하였으며, 부정류 상태에서 Manning 조도계수의 시간적, 공간적 변동성을 고려하여 모형의 보정을 수행하였다. 이어지는 오차보정 과정에서는 장단기 기억 신경망(LSTM) 모형을 사용하여 동수역학적 모형의 시스템적 오차를 찾아내고자 하였다. 모형의 정확성을 평가하기 위하여 동수역학적 모형, 두 개의 LSTM 기반 모형(LSTM1, LSTM2) 및 혼합모형 등 4개 모형의 성능을 비교하였다. LSTM1은 현재까지의 관측 자료들만을 입력변수로 사용하는 모형이며, LSTM2는 여기에 이후 시간에서의 동수역학적 모형 경계조건들을 입력변수로 추가하여 수위를 예측하도록 수립된 모형이다. 평균제곱근 오차를 평가 지표로 하여 모형들의 성능을 평가한 결과, 혼합모형이 개별 모형들에 비하여 홍수위의 예측 정확성을 현저하게 개선하는 것으로 나타났다. 이는 혼합모형이 동수역학적 모형과 LSTM 모형의 상호 보완적인 장점들을 모두 취함으로써 가능한 것이라 할 수 있다. 또한 LSTM 기반의 혼합모형은 인공신경망 기반의 혼합모형보다도 우수한 성능을 보여, 예측 선행시간이 3시간인 경우에도 제곱평균제곱근 오차 10 cm 이내로 홍수위를 예측하는 것으로 나타났다.

A hybrid approach combining a distributed parameter hydrodynamic flow model and a deep learning based error correction model is proposed to enhance the forecasting accuracy of flood water levels. The hydrodynamic model for the downstream reach of the Han River was calibrated by considering the spatial distribution and temporal variation of Manning's roughness coefficient in unsteady flow conditions. In the following error correction procedure, the systematic errors of the hydrodynamic model were captured using long short term memory (LSTM) model. To assess the model accuracy, performances of the hydrodynamic model, two LSTM based models (LSTM1 and LSTM2), and the hybrid model were compared. Regarding inputs, LSTM1 considers the historical observations only, and LSTM2 considers not only the historical observations that used in LSTM1 but also the prescribed boundary conditions required for the hydrodynamic forecast model. The root-mean-square errors were used for the evaluation of model performance. The hybrid model significantly improved the forecasting accuracy of flood water levels compared to individual models, which indicates that the hybrid model is able to take advantage of complementary strengths of both the hydrodynamic model and the LSTM model. The LSTM-based hybrid model also outperformed the hybrid model based on artificial neural network, resulting in less than 10 cm of root-mean-square error for 3-hour lead time forecasting of the flood water level.

본 연구의 목적은 사회서비스 수요를 구성하는 이용자 수와 제공기관 수를 예측 할 수 있도록 시계열 모형을 활용하여 각각의 예측 값을 구성하고, 실제 관측된 값과의 차이를 확인하여 사회서비스분야에서 시계열 예측모형의 타당성을 검증하는 연구이다. 분석 자료는 한국보건복지정보개발원에서 발간한 사회서비스 제공기관 공급실태분석에서 제시된 발달재활서비스 이용 현황을 연구 목적에 따라 가공하여 이차 분석하였다. 분석결과 이용자 수는 ARIMA(1,1,0) 모형이, 제공기관 수는 ARIMA(0,1,1) 모형이 최적의 예측모형으로 제시되었다. 예측모형에 의한 예측 값은 관측 값과의 어느 정도 차이는 있었지만, 관측값은 예측값의 최대값과 최소값의 범위에 놓여 있었다. 따라서 사회서비스의 이용자를 활용한 수요예측과 제공기관을 활용한 공급예측의 모형구축에 대한 타당성은 가능할 수 있음을 확인할 수 있었다.

The primary goal of the study is to examine the possibility of applying the time series model to forecasting demand and supply of social services. In the study, we used survey data based on a nationally represented sample which is secondary processed data. We selected developmental rehabilitation service. The analysis, we made models of a demand and a supply using time series analysis. Utilizing the estimates, we identified each model's pattern. This study provides an empirical evidence to suggest benefits of using the time series model for forecasting the demand and the supply pattern of newly introduced social services. We also provide discussions on policy implications of utilizing demand and supply time series models in the process of developing new social services.