• Proceedings of the Korea Water Resources Association Conference
• /
• 2017.05a
• /
• pp.162-162
• /
• 2017

최근 우리나라는 전 지구적인 기후변화로 인하여 집중호우 및 돌발 홍수와 같은 극치 사상들이 증가하고 있는 추세이며, 이에 대한 분석을 위해 극치 분포를 이용한 수문통계적 특성에 대한 접근이 주로 이루어지고 있다. 이를 위해서는 충분한 수의 자료가 필요하나 우리나라 강우자료는 지점별로 자료 보유 년 수가 비교적 많지 않기 때문에, 이러한 문제를 극복하기 위하여 하나의 지역, 즉 주어진 지점을 포함하여 수문학적으로 동일한 조건을 만족하는 주변 지점의 자료를 모두 포함하여 빈도해석을 실시하는 지역빈도해석이 필요하다. 따라서 본 연구에서는 지역빈도해석과 두 개의 형상매개변수를 포함하여 다양한 극치 수문통계특성을 나타낼 수 있다고 알려진 Burr XII 분포를 이용하여 우리나라 강우자료에 대한 그 적용성을 살펴보았다. 이를 위해 군집분석을 통한 강우지점의 지역화 과정을 거치고 분류된 지역을 L-moment ratio diagram에 도시하여, Burr XII 분포 영역 내 포함여부를 통해 Burr XII 분포의 적합도를 도시적으로 살펴보고, Hosking and Wallis (1997)이 제안한 적합성 척도($^{IST}$)를 통한 적합성 여부를 판별하였다. 또한 우리나라 강우자료에 비교적 적합하다고 알려진 분포인 generalized extreme value, generalized logistic, Gumbel 분포와의 비교를 위해, 전체 지역에 대하여 재현기간에 따른 상대편의 (relative bias)와 상대평균제곱근오차 (relative root mean square error)를 산정하여 Burr XII 분포형의 적용 가능성을 살펴보았다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2016.05a
• /
• pp.158-158
• /
• 2016

본 연구에서는 극치 분포의 오른쪽 꼬리 부분 예측 시 안정적인 확률수문량 산정하는 확률분포형과 매개변수 추정 방법을 평가하기 위해 Monte Carlo 모의를 수행하였다. 수문자료의 빈도해석에 적합한 것으로 알려진 generalized extreme value (GEV), Gumbel (GUM), generalized logistic (GLO), gamma3 (GAM3), normal (NOR), log-normal3 (LN3) 총 6개의 확률분포형을 바탕으로 오른쪽 꼬리 부분의 확률수문량 추정 성능을 모의 실험을 통해 평가하고자 한다. 30년 이상 자료를 보유한 기상청 지점의 지속기간별 연최대값 자료를 분석한 결과를 바탕으로 모분포를 GEV분포로 선정하였으며 평균이 1.0, 표준편차 0.5, 왜곡도 계수는 0.5, 1.0, 2.0, 3.0, 4.0이 되도록 가정하였다. 또한 자료 길이에 따른 성능 평가를 위해 표본 크기 20, 50, 100, 150, 200개에 대해 분석을 수행하였다. 위와 같은 가정으로 총 25종류(왜곡도계수 5개 ${\times}$ 표본 크기 5개)의 발생된 모분포에 6가지의 확률분포형과 3가지의 매개변수 추정방법(모멘트법, 최우도법, 확률가중모멘트법)을 조합한 18가지의 모델을 비교 분석해보았다. 평가방법으로는 평균 제곱근 오차(Root Mean Square Error, RMSE), 편의(bias), 평균 상대오차(Mean Relative Difference, MRD), 평균 절대 상대오차(Mean Absolute Relative Difference, MARD)를 사용하여 적용 모델의 성능을 비교 분석하였다.

• Advances in Computational Design
• /
• v.7 no.3
• /
• pp.253-279
• /
• 2022

This study investigated and predicted the Marshall stability of glass-fiber asphalt mix, carbon-fiber asphalt mix and glass-carbon-fiber asphalt (hybrid) mix by using machine learning techniques such as Artificial Neural Network (ANN), Support Vector Machine (SVM) and Random Forest(RF), The data was obtained from the experiments and the research articles. Assessment of results indicated that performance of the Artificial Neural Network (ANN) based model outperformed applied models in training and testing datasets with values of indices as; coefficient of correlation (CC) 0.8492 and 0.8234, mean absolute error (MAE) 2.0999 and 2.5408, root mean squared error (RMSE) 2.8541 and 3.3165, relative absolute error (RAE) 48.16% and 54.05%, relative squared error (RRSE) 53.14% and 57.39%, Willmott's index (WI) 0.7490 and 0.7011, Scattering index (SI) 0.4134 and 0.3702 and BIAS 0.3020 and 0.4300 for both training and testing stages respectively. The Taylor diagram also confirms that the ANN-based model outperforms the other models. Results of sensitivity analysis show that Carbon fiber has a major influence in predicting the Marshall stability. However, the carbon fiber (CF) followed by glass-carbon fiber (50GF:50CF) and the optimal combination CF + (50GF:50CF) are found to be most sensitive in predicting the Marshall stability of fibrous asphalt concrete.

• Nuclear Engineering and Technology
• /
• v.54 no.4
• /
• pp.1439-1448
• /
• 2022

Background: We investigated the feasibility of in vitro radiosensitivity prediction with gene expression using deep learning. Methods: A microarray gene expression of the National Cancer Institute-60 (NCI-60) panel was acquired from the Gene Expression Omnibus. The clonogenic surviving fractions at an absorbed dose of 2 Gy (SF2) from previous publications were used to measure in vitro radiosensitivity. The radiosensitivity prediction model was based on the convolutional neural network. The 6-fold cross-validation (CV) was applied to train and validate the model. Then, the leave-one-out cross-validation (LOOCV) was applied by using the large-errored samples as a validation set, to determine whether the error was from the high bias of the folded CV. The criteria for correct prediction were defined as an absolute error<0.01 or a relative error<10%. Results: Of the 174 triplicated samples of NCI-60, 171 samples were correctly predicted with the folded CV. Through an additional LOOCV, one more sample was correctly predicted, representing a prediction accuracy of 98.85% (172 out of 174 samples). The average relative error and absolute errors of 172 correctly predicted samples were 1.351±1.875% and 0.00596±0.00638, respectively. Conclusion: We demonstrated the feasibility of a deep learning-based in vitro radiosensitivity prediction using gene expression.

• Atmosphere
• /
• v.22 no.1
• /
• pp.13-28
• /
• 2012

The $PM_{10}$ concentration data is useful for indentifying intensity and a transport way of Asian dust. However, it is difficult to identify them properly due to the limited spatial resolution and coverage. Therefore, a methodology to estimate $PM_{10}$ concentration using visibility data obtained from synoptic observation was developed. To derive the converting function, correlation between visibility and $PM_{10}$ concentration is investigated using visibility and $PM_{10}$ concentration data observed at 20 stations in Korea from 2005 to 2009. To minimize bias due to atmospheric moisture, data with higher relative humidity over a critical value were eliminated while deriving $PM_{10}$-visibility relationship. As a result, an exponentially decreasing function of visibility is obtained under the condition that relative humidity is less than 82%. Verification of the visibility converting function to $PM_{10}$ concentration was carried out for the dust cases in 2010. It was found that spatial distributions of $PM_{10}$ calculated by visibility are in good agreement with the observed $PM_{10}$ distribution, especially for the strong dust cases in 2010. And correlation between the derived and observed $PM_{10}$ concentration was 0.63. We applied the function to obtain distributions of $PM_{10}$ concentration over North Korea, in which concentration data are not available, and compared them with satellite derived dust index, IODI distributions for dust cases in 2010. It is shown that the visibility function estimates quite similar patterns of dust concentration with IODI image, which suggests that it can contribute for prediction by indentifying transport route of Asian dust.

• Magazine of the Korean Society of Agricultural Engineers
• /
• v.45 no.5
• /
• pp.97-109
• /
• 2003

This study is mainly conducted to derive the design drought rainfall by the consecutive duration using probability weighted moments with rainfall in the regional drought frequency analysis. It is anticipated to suggest optimal design drought rainfall of hydraulic structures for the water requirement and drought frequency of occurrence for the safety of water utilization through this study. Preferentially, this study was conducted to derive the optimal regionalization of the precipitation data that can be classified by the climatologically and geographically homogeneous regions all over the regions except Cheju and Ulreung islands in Korea. Five homogeneous regions in view of topographical and climatological aspects were accomplished by K-means clustering method. Using the L-moment ratio diagram and Kolmogorov-Smirnov test, generalized extreme value distribution was confirmed as the best fitting one among applied distributions. At-site and regional parameters of the generalized extreme value distribution were estimated by the method of L-moments. Design drought rainfalls using L-moments following the consecutive duration were derived by the at-site and regional analysis using the observed and simulated data resulted from Monte Carlo techniques. Relative root-mean-square error (RRMSE), relative bias (RBIAS) and relative reduction (RR) in RRMSE for the design drought rainfall derived by at-site and regional analysis in the observed an simulated data were computed and compared. In has shown that the regional frequency analysis procedure can substantially more reduce the RRMSE. RBIAS and RR in RRMSE than those of at-site analysis in the prediction of design drought rainfall. Consequently, optimal design drought rainfalls following the regions and consecutive durations were derived by the regional frequency analysis.

• The Korean Journal of Applied Statistics
• /
• v.26 no.6
• /
• pp.1033-1041
• /
• 2013

In this paper, we investigate several variance estimators for pps systematic sampling. Unfortunately, there is no unbiased variance estimators for a systematic sample because systematic sampling can be regarded as a random selection of one cluster. This study provides guidance on which variance estimator may be more appropriate than others in several circumstances. We judge the efficiency of variance estimators for systematic sampling based on of their relative biases and relative mean square error. Also, we investigate variance estimation problems for two-stage systematic sampling applied for the Food Raw Material Consumption Survey and the Establishment Labor Force Survey simulation study, in order to consider the popular two-stage pps systematic sample design for establishment and household survey in Korea.

• Water Engineering Research
• /
• v.5 no.4
• /
• pp.185-193
• /
• 2004

Estimates of annual actual evapotranspiration are needed in water balance studies, water resources management projects, and many different types of hydrologic studies. This study validated a set of 5 empirical equations of estimating annual actual evapotranspiration with climate data on 11 watersheds, and evaluated the further applicability of these forms in estimating annual runoff on watershed level. Five empirical equations generally overestimated annual evapotranspiration, with relative errors ranging $3.3\%$ to $47.2\%$. The results show that Schreiber formula can be applicable in determining annual evapotranspiration in sub-humid region that is classified by aridity index, while Zhang equation gave better results than the remaining methods in humid region. The mean differences for annual evapotranspiration bias over 11 watersheds are Zhang, Schreiber, Budyko, Pike, and Ol'dekop formula from lowest to highest. The empirical equations provide a practical tool to help water resources managers in estimating regional water resources on ungauged large watershed.

• Proceedings of the KIEE Conference
• /
• 1995.07c
• /
• pp.1434-1436
• /
• 1995

We designed and fabricated a highly sensitive magnetotransistor which employes the emitter region as a Hall plate for inducing Hall voltage across the emitter. The Hall voltage modulates the emitter basic junction bias on both sides of the emitter so that a large collector current difference is resulted. The specially designed $p^+$ ring around the emitter enhances accumulation of drifted electrons in the emitter and thus the Hall voltage. A relative sensitivity of 240/tesla is measured by operating the device in the saturation mode.

• Journal of Korean Institute of Industrial Engineers
• /
• v.29 no.2
• /
• pp.114-125
• /
• 2003

This paper proposes multivariate process capability indices (PCIs) for skewed populations using $T^2$rand modified process region approaches. The proposed methods are based on the multivariate version of a weighted standard deviation method which adjusts the variance-covariance matrix of quality characteristics and approximates the probability density function using several multivariate Journal distributions with the adjusted variance-covariance matrix. Performance of the proposed PCIs is investigated using Monte Carlo simulation, and finite sample properties of the estimators are studied by means of relative bias and mean square error.