• The Korean Journal of Applied Statistics
• /
• v.15 no.1
• /
• pp.139-151
• /
• 2002

This study is concerned with model selection and diagnostics for nonlinear regression model through Bayes factor. In this paper, we use informative prior and simulate observations from the posterior distribution via Markov chain Monte Carlo. We propose the Laplace approximation method and apply the Laplace-Metropolis estimator to solve the computational difficulty of Bayes factor.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2004.05b
• /
• pp.719-723
• /
• 2004

본 연구에서는 일 최고, 최소치 유출량 계열을 확충하기 위해 ARIMA(p,d,q) 모형을 이용하였으며, 분석 자료의 경향성 유무를 파악하기 위해 Mann-Kendal 비모수적 검정을 실시하였다. 분석 결과, 최고 최소 유출량 자료 모두 경향성이 없는 것으로 분석되었다. ARIMA(p,d,q) 모형의 최적 차수를 결정하기 위해 ACF, PACF, AIC, 그리고 SBC(Schwarz Bayesian Criterion) 검사를 실시하였으며 이를 통해 최적의 ARMA 모형을 결정하였다. 일 최대치 자료의 경우 추계학적 경향 보다는 무작위적 특성을 보였으며, 일 최소치 자료계열 경우, ARMA(1,0) 모형이 최적 모형으로 선정되었다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2000.11a
• /
• pp.387-392
• /
• 2000

전자 상거래에서 고객의 성향을 이해하기 위해서는 일반적으로 판매 실무에서의 경험과 전문적인 지식을 필요로 하게 된다. 데이터 마이닝은 고객들에 대한 데이터의 분석을 통해서 이러한 성향들을 알아내는 것을 목표로 한다. 베이지안 네트워크는 DAG(Directed Acyclic Graph)를 이용하여 데이터의 구조를 시각적으로 표현하여 주는 확률모형으로 변수사이의 종속관계를 밝히고 데이터 마이닝의 기법으로 이용할 수 있다. 본 논문에서는 베이지안 네트워크를 사용하여 전자 상거래 고객들의 성향을 분석하기 위한 방법을 제시한다. 또한 고객성향에 대한 주요 요인을 분석하기 위해 Descriminant 모형을 이용하고 그 유용성을 다른 방법들과 비교하였다.

• Journal of Korea Water Resources Association
• /
• v.44 no.5
• /
• pp.339-350
• /
• 2011

This study developed a climate informed Bayesian nonstationary frequency model which allows us to forecast seasonal summer rainfall at Nakdong River. We constructed a 37-year summer rainfall data set from 10 weather stations within Nakdong river basin, and two climate indices from sea surface temperature (SST) and outgoing longwave radiation (OLR) were derived through correlation analysis. The selected SST and OLR have been widely acknowledged as a climate driver for summer rainfall. The developed model was applied first to the 2010-year summer rainfall (888.1 mm) in order to assure ourself. We demonstrated model performance by comparing posterior distributions. It was confirmed that the proposed model is able to produce a reasonable forecast. The forecasted value is about 858.2 mm, and the difference between forecast and observation is about 30 mm. As the second case study, 2011-year summer rainfall forecast was made using an observed winter SSTs and an assumed 50% value of OLRs. The forecasted value is 967.7 mm and associated exceedance probability over average summer rainfall 680 mm is 92.9%. In addition, 50-year return period for summer rainfall was projected through the nonstationary frequency model. An exceedance probability over 1,400 mm corresponding to the 50-year return level is about 73.7%.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.307-307
• /
• 2021

개념적 강우-유출모형에서 토양수분과 관련된 물리적 거동은 간략화 된 형태로 강우 및 온도자료를 활용하여 중간변량(state variable)으로 간접적으로 고려되고 있다. 특히 강우-유출모형에 초기함수 조건은 선행함수조건을 고려하여 수문지질학적 평가를 통하여 결정되어야 하나, 일반적으로 가정되거나 모형에서 간략화 된 분석과정을 통해 추정되고 있다. 본 연구에서는 토양의 Water Balance 모형 기반의 개념적 토양수분 추정모형을 활용하였다. 토양수분의 시간적 변동성을 평가하는데 있어서 연속적으로 측정된 In-situ 토양수분 자료를 이용하여 모형의 적합성을 평가하였다. Green-Ampt 방법과 중력식 침투방법과 온도를 활용한 증발산 추정기법을 연계한 토양함수 평가 모형을 개발하였다. In-situ 토양수분 자료와 유역의 강수량 및 온도자료를 이용한 관련 매개변수를 Bayesian 기법을 통해 추정하였으며 매개변수의 민감도를 평가하여 제시하였다. 최종적으로 제안된 모형의 활용측면에서 강우-유출모형의 초기함수 조건으로써의 역할을 평가하였다. 구체적으로 첨두유량 및 유출고와 초기함수조건과의 관계를 제시하고 강우-유출모형에서 활용방안을 제시하고자 한다.

• Journal of Korea Water Resources Association
• /
• v.54 no.10
• /
• pp.747-757
• /
• 2021

Quantification of extreme rainfall is very important in establishing a flood protection plan, and a general measure of extreme rainfall is expressed as an T-year return level. In this study, a method was proposed for quantifying spatial distribution and uncertainty of daily rainfall depths with various return periods using a hierarchical Bayesian model combined with climate and geographical information, and was applied to the Seoul-Incheon-Gyeonggi region. The annual maximum daily rainfall depth of six automated synoptic observing system weather stations of the Korea Meteorological Administration in the study area was fitted to the generalized extreme value distribution. The applicability and reliability of the proposed method were investigated by comparing daily rainfall quantiles for various return levels derived from the at-site frequency analysis and the regional frequency analysis based on the index flood method. The uncertainty of the regional frequency analysis based on the index flood method was found to be the greatest at all stations and all return levels, and it was confirmed that the reliability of the regional frequency analysis based on the hierarchical Bayesian model was the highest. The proposed method can be used to generate the rainfall quantile maps for various return levels in the Seoul-Incheon-Gyeonggi region and other regions with similar spatial sizes.

• Journal of the military operations research society of Korea
• /
• v.33 no.2
• /
• pp.101-113
• /
• 2007

The period of military personnel service will be phased down by 2014 according to 'The law of National Defense Reformation' issued by the Ministry of National Defense. For this reason, the ROK army provides discrimination education to 'newly recruited privates' for more effective individual performance in the on-the-job training. For the training to be more effective, it would be essential to predict the degree of achievements by new privates in the training. Thus, we used data mining techniques to develop a classification model which classifies the new privates into one of two achievements groups, so that different skills of education are applied to each group. The target variable for this model is a binary variable, whose value can be either 'a group of general control' or 'a group of special control'. We developed four pure classification models using Neural Network, Decision Tree, Support Vector Machine and Naive Bayesian. We also built four hybrid models, each of which combines k-means clustering algorithm with one of these four mining technique. Experimental results demonstrated that the highest performance model was the hybrid model of k-means and Neural Network. We expect that various military education programs could be supported by these classification models for better educational performance.

• Communications for Statistical Applications and Methods
• /
• v.1 no.1
• /
• pp.94-101
• /
• 1994

본 논문은 패널(panel) 자기회귀 모형에서 자기회귀 계수의 추정을 베이지안 방법으로 접근하였는데, 이 때 특별히 Gibbs Sampling 방법을 이용하여 사후분포를 계산하였다. 또한 모의 실험을 통하여 자기회귀계수를 Gibbs Sampling 방법으로 추정한 베이지안 추정치가 non-Bayesian 방법으로 구한 추정치보다 더 우월함을 보였다.

• Journal of Korean Society on Water Environment
• /
• v.36 no.4
• /
• pp.300-313
• /
• 2020

The Bayesian approach can be used to estimate hydrologic model parameters from the prior expert knowledge about the parameter values and the observed data. The purpose of this study was to compare the performance of the two Bayesian methods, the Metropolis-Hastings (MH) algorithm and the Generalized Likelihood Uncertainty Estimation (GLUE) method. These two methods were applied to the TANK model, a hydrological model comprising 13 parameters, to examine the uncertainty of the parameters of the model. The TANK model comprises a combination of multiple reservoir-type virtual vessels with orifice-type outlets and implements a common major hydrological process using the runoff calculations that convert the rainfall to the flow. As a result of the application to the Nam River A watershed, the two Bayesian methods yielded similar flow simulation results even though the parameter estimates obtained by the two methods were of somewhat different values. Both methods ensure the model's prediction accuracy even when the observed flow data available for parameter estimation is limited. However, the prediction accuracy of the model using the MH algorithm yielded slightly better results than that of the GLUE method. The flow duration curve calculated using the limited observed flow data showed that the marginal reliability is secured from the perspective of practical application.

• Journal of Korea Water Resources Association
• /
• v.48 no.10
• /
• pp.781-791
• /
• 2015

Dam risk analysis requires a systematic process to ensure that hydrologic variables (e.g. precipitation, discharge and water surface level) contribute to each other. However, the existing dam risk approach showed a limitation in assessing the interdependencies across the variables. This study aimed to develop Bayesian network based dam risk analysis model to better characterize the interdependencies. It was found that the proposed model provided advantages which would enable to better identify and understand the interdependencies and uncertainties over dam risk analysis. The proposed model also provided a scenario-based risk evaluation framework which is a function of the failure probability and the consequence. This tool would give dam manager a framework for prioritizing risks more effectively.