• Proceedings of the Korea Water Resources Association Conference
• /
• 2010.05a
• /
• pp.1164-1168
• /
• 2010

수자원이 우리 생활의 전반적으로 중요한 역할을 차지하면서 댐의 효율적인 운영과 안정적인 용수공급에 대한 연구는 지속적으로 수행되어지고 있다. 1990년대 이후 비선형적인 특성을 잘 모의하는 장점을 가진 인공신경망(ANN)을 이용하여 유입량 예측에 대한 많은 연구가 수행되었다. 하지만 ANN 모형을 포함한 회귀모형은 월 강우 및 유입량의 예측에 대해 간편하게 사용을 할 수 있지만, 예측의 정확성에 한계를 가지고 있다. 본 연구에서는 ANN 모형과 회귀모형의 예측오차를 후처리 과정을 통하여 오차를 줄임으로써 예측모형의 성과를 향상시키는 방법을 제안하였다. 연구지역은 금강수계의 대청댐 유역으로, 1982년 9월부터 2005년 12월에 해당하는 유역 내 11개 지점의 강우관측소에서 관측한 월 강우와 댐 유입량을 수집하여 모형을 구축하였다. 강우량과 유입량 자료에 대해 자기상관함수와 교차상관함수를 이용하여 입력변수를 결정하였고, 정규화를 통한 전처리 과정을 거쳐 ANN 모형과 회귀모형을 이용한 예측모형을 구축하였으며, 예측성과의 향상을 위하여 군집 분석을 이용하여 오차를 재조정하였다. 이러한 오차 후처리 과정을 포함한 모형은 RMSE와 상관계수를 이용하여 비교 평가한 결과, 예측성과를 약 40% 정도 향상시켰다.

• Journal of the Korean Data and Information Science Society
• /
• v.23 no.1
• /
• pp.71-77
• /
• 2012

This paper introduces the projection methods by nonlinear regression model. To predict the student numbers, a log model and an involution model as the kind of a trend-extrapolation method are used. Empirical evidence shows that a projection by log model is better than by involution model with the confidence interval estimations for the coefficients of determination.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2018.05a
• /
• pp.23-23
• /
• 2018

본 연구에서는 다중분위회귀분석모형(Multiple Quantile Regression Model, MQRM)과 MODIS(MODerate resolution Imaging Spectroradiometer) LST (Land Surface Temperature) 자료를 이용하여 전국 공간토양수분을 산정하였다. 공간토양수분을 산정하기 위한 과정은 크게 두가지로 구분된다. 첫 번째로 기존의 MODIS LST 자료를 조건부 합성 보정기법을 적용하여 실측 LST 자료와 비교하여 위성 LST 자료가 갖고 있는 오차를 보정하였다. 그 결과, 조건부 합성 보정기법을 적용하기전 전국 71개 지상관측지점에서 관측한 실측 LST와 MODIS LST의 $R^2$는 전체 평균 0.70으로 어는정도 유의성 있는 상관관계를 나타냈으나 조건부 합성 보정기법을 적용한 후 실측 LST와 MODIS LST의 $R^2$는 전체 평균 0.92로 상당히 크게 향상됨을 알 수 있었다. 두 번째로 보정된 MODIS LST를 이용하여 다중분위회귀분석 모형을 개발하고 토양수분을 예측하는 단계로 입력자료로 위성영상 자료와 관측자료를 융합하여 사용하였다. 위성영상 자료로는 보정된 MODIS LST와 MODIS NDV를 구축하였고 일단위 강수량 및 일조시간의 기상자료는 기상청으로부터 전국 71개 지점에 대해 구축하여 IDW 공간보간기법을 이용한 공간자료로 구축하였다. 토양수분 결과를 비교하기 위한 관측 토양수분은 자동농업기상관측(Automated Agriculture Observing System, AAOS)지점에서 2013년 1월부터 2015년 12월까지의 실측 일단위 토양수분 자료를 구축하여 사용하였다. 다중분위회귀분석 모형은 LST 인자를 중심으로 각각의 분위(0.05, 0.25, 0.5, 0.75, 0.95)에 해당되는 값의 회귀식을 NDVI, 강수 입력자료를 독립인자로서 조합하여 계절 및 토성에 따른 총 80개의 회귀식을 산정하였다. 관측 토양수분과 모의 토양수분을 비교한 결과 $R^2$가 0.70 (철원), 0.90 (춘천), 0.85 (수원), 0.65 (서산), 0.78 (청주), 0.82 (전주), 0.62 (순천), 0.63 (진주), 0.78 (보성)로 높은 상관성을 보였다. 본 연구에서는 다중분위회귀 모형의 성능을 검증하기 위해 기존의 다중선형회귀모형의 결과와 비교하여 크게 개선됨을 나타냈다.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.1
• /
• pp.141-150
• /
• 2015

Graphical paradigms for assessing the adequacy of models in logistic regression are discussed. The residual plot has been widely used as a graphical tool for evaluating the adequacy of the model. However, this approach works well only for linear models with constant variance, and the alternative approach, the marginal model plot, has its defects as well. We suggest a Chi-residual plot that overcomes the potential shortcomings of the marginal model plot.

• Journal of the Korean Data and Information Science Society
• /
• v.23 no.6
• /
• pp.1117-1125
• /
• 2012

The occurrence of food poisoning is usually modeled by meteorological variables like the temperature and the humidity. In this paper, we investigate the relationship between food poisoning occurrence and climate variables in Korea and compare Poisson regression and autoregressive moving average model to select the forecast model. We confirm that lagged climate variables affect the food poisoning occurrences. However, it turns out that, from the viewpoint of the prediction, the number of previous occurrences is more influential to the current occurrence than meteorological variables and Poisson regression model is less reliable.

• Journal of the Korean Data and Information Science Society
• /
• v.20 no.2
• /
• pp.293-299
• /
• 2009

In this article, we consider linear models such as regression, ARIMA (autoregressive integrated moving average), and regression+ARIMA (regression with ARIMA errors) for predicting hourly ozone concentration level in two areas of Daegu. Based on RASE(root average squared error), it is shown that the ARIMA is the best model in one area and that the regression+ARIMA model is the best in the other area. We further analyze the residuals from the optimal models, so that we might predict the ozone warning days where at least one of the hourly ozone concentration levels is over 120 ppb. Based on the training data in the years from 2000 to 2003, it is found that 35 ppb is a good cutoff value of residulas for predicting the ozone warning days. In on area of Daegu, our method predicts correctly one of two ozone warning days of 2004 as well as all of the remaining 364 non-warning days. In the other area, our methods predicts correctly all of one ozone warning days and 365 non-warning days of 2004.

• Journal of the Korean Data and Information Science Society
• /
• v.21 no.5
• /
• pp.831-839
• /
• 2010

Least squares support vector machine (LS-SVM) is a kernel trick gaining a lot of popularities in the regression and classification problems. We use LS-SVM to propose a iterative algorithm for a nonlinear generalized autoregressive conditional heteroscedasticity model in the mean (GARCH-M) model to estimate the mean and the conditional volatility of stock market returns. The proposed method combines a weighted LS-SVM for the mean and unweighted LS-SVM for the conditional volatility. In this paper, we show that nonlinear GARCH-M models have a higher performance than the linear GARCH model and the linear GARCH-M model via real data estimations.

• The Korean Journal of Applied Statistics
• /
• v.18 no.3
• /
• pp.701-712
• /
• 2005

The suppression for logistic regression models has been debated no longer than that for linear regression models since, among many other reasons, sum of squares for regression (SSR) or coefficient of determination ($R^2$) could be defined into various ways. Based on four kinds of $R^2$'s: two kinds are most preferred, and the other two are proposed by Liao & McGee (2003), four kinds of SSR's are derived so that the suppression for logistic models is explained. Many data fitted to logistic models are generated by Monte Carlo method. We explore when suppression happens, and compare with that for linear regression models.

• Journal of the Korean Geographical Society
• /
• v.45 no.6
• /
• pp.806-819
• /
• 2010

Population data has been essential and fundamental in spatial analysis and commonly aggregated into political boundaries. A conventional method for population distribution estimation was a regression model with land use data, but the estimation process has limitation because of spatial autocorrelation of the population data. This study aimed to improve the accuracy of population distribution estimation by adopting a Regression-Kriging method, namely RK Model, which combines a regression model with Kriging for the residuals. RK Model was applied to a part of Seoul metropolitan area to estimate population distribution based on the residential zones. Comparative results of regression model and RK model using RMSE, MAE, and G statistics revealed that RK model could substantially improve the accuracy of population distribution. It is expected that RK model could be adopted actively for further population distribution estimation.

• Journal of the Korean Data and Information Science Society
• /
• v.25 no.6
• /
• pp.1491-1498
• /
• 2014

The major criticism of optimal experimental design is that it depends heavily on the model and its accompanying assumption that often leads the number of support points equal to the number of parameters in the model. Often in the past, a polynomial model of higher degree is assumed to handle the experimental design for the polynomial regression of lower degree. In this paper we searched the possible set of designs which are robust to the departure of the assumed model. The designs are categorized with respect to D-efficiency. The approach by O'Brien (1995) was discussed in univariate polynomial regression model setting.