Variance Function Estimation using SVM and Jack-Knife Method

심주용,조대현 한국자료분석학회 2011 Journal of the Korean Data Analysis Society Vol.13 No.5

We propose an estimation method of variance functions in the heteroscedastic regression. To obtain more stable variance function estimation we use the averages of squared jack-knifed residuals obtained from support vector machine. We use the iteratively reweighted least squares procedure to estimate the regression function in support vector machine for easy model selection method which employs the generalized approximate cross validation techniques for choosing the hyper-parameters which affect the performance of the regression method. An artificial example and a real example are provided to indicate the usefulness of the proposed method for the estimation of variance functions.

소지역 실업률의 패널추정을 위한 일반화커널추정방정식

심주용,김영원,황창하 한국데이터정보과학회 2013 한국데이터정보과학회지 Vol.24 No.6

The high unemployment rate is one of the major problems in most countries nowadays. Hence, the demand for small area labor statistics has rapidly increased over the past few years. However, since sample surveys for producing official statistics are mainly designed for large areas, it is difficult to produce reliable statistics at the small area level due to small sample sizes. Most of existing studies about the small area estimation are related with the estimation of parameters based on cross-sectional data. By the way, since many official statistics are repeatedly collected at a regular interval of time, for instance, monthly, quarterly, or yearly, we need an alternative model which can handle this type of panel data. In this paper, we derive the generalized kernel estimating equation which can model time-dependency among response variables and handle repeated measurement or panel data. We compare the proposed estimating equation with the generalized linear model and the generalized estimating equation through simulation, and apply it to estimating the unemployment rates of 25 areas in Gyeongsangnam-do and Ulsan for 2005. 오늘날 높은 실업률은 대부분의 국가에서 중요한 문제 중의 하나이다. 한편 소지역의 노동 관련통계에 대한 요구가 지난 몇년간 급속도로 증가하였다. 그러나 대부분의 공식통계를 생산하기 위한표본설계는 대영역의 통계를 생산할 목적으로 설계되기 때문에 소지역의 경우 배정되는 표본조사단위수가 극히 적어 신뢰성 있는 통계 산출이 어렵다. 그리고 소지역 추정에 대한 대부분의 기존 연구들은 특정 시점에서의 추정에 국한 되어 왔다. 그러나 대부분의 공식통계들은 월, 분기 또는 연 단위로 측정되는 패널자료이기 때문에 이를 고려한 추정방법이 필요하다. 본 논문에서는 패널자료의 분석을 위해 유용하게 사용되고 있는 일반화추정방정식의 비모수적 버전인 일반화커널추정방정식을 도출하여 조사시점을 고려한 소지역 실업률의 추정에 활용하는 방안을 제안한다. 모의실험을 통하여 일반화커널추정방정식 방법, 일반화추정방정식 방법 및 일반화선형모형과 비교한다. 그리고 2005년 1월부터 12월까지 경상남도 및 울산광역시의 25개 시군구의 경제활동인구조사의 패널자료에 위에서 언급한 세 가지 방법을 적용하여 해당 소지역의 월별 실업률을 추정한다.

Forecasting volatility via conditional autoregressive value at risk model based on support vector quantile regression

심주용,황창하 한국데이터정보과학회 2011 한국데이터정보과학회지 Vol.22 No.3

The conditional autoregressive value at risk (CAViaR) model is useful for risk management, which does not require the assumption that the conditional distribution does not vary over time but the volatility does. But it does not provide volatility forecasts, which are needed for several important applications such as option pricing and portfolio management. For a variety of probability distributions, it is known that there is a constant relationship between the standard deviation and the distance between symmetric quantiles in the tails of the distribution. This inspires us to use a support vector quantile regression (SVQR) for volatility forecasts with the distance between CAViaR forecasts of symmetric quantiles. Simulated example and real example are provided to indicate the usefulness of proposed forecasting method for volatility.

SVQR with asymmetric quadratic loss function

심주용,김말숙,석경하 한국데이터정보과학회 2015 한국데이터정보과학회지 Vol.26 No.6

Support vector quantile regression (SVQR) can be obtained by applying support vector machine with a check function instead of an e-insensitive loss function into the quantile regression, which still requires to solve a quadratic program (QP) problem which is time and memory expensive. In this paper we propose an SVQR whose objective function is composed of an asymmetric quadratic loss function. The proposed method overcomes the weak point of the SVQR with the check function. We use the iterative procedure to solve the objective problem. Furthermore, we introduce the generalized cross validation function to select the hyper-parameters which affect the performance of SVQR. Experimental results are then presented, which illustrate the performance of proposed SVQR.

Expected shortfall estimation using kernel machines

심주용,황창하 한국데이터정보과학회 2013 한국데이터정보과학회지 Vol.24 No.3

In this paper we study four kernel machines for estimating expected shortfall, which are constructed through combinations of support vector quantile regression (SVQR),restricted SVQR (RSVQR), least squares support vector machine (LS-SVM) and sup-port vector expectile regression (SVER). These kernel machines have obvious advan-tages such that they achieve nonlinear model but they do not require the explicit form of nonlinear mapping function. Moreover they need no assumption about the underly-ing probability distribution of errors. Through numerical studies on two artificial and two real data sets we show their effectiveness on the estimation performance at various confidence levels.

Estimating multiplicative competitive interaction model using kernel machine technique

심주용,김말숙,박혜정 한국데이터정보과학회 2012 한국데이터정보과학회지 Vol.23 No.4

We propose a novel way of forecasting the market shares of several brands simultaneously in a multiplicative competitive interaction model, which uses kernel regression technique incorporated with kernel machine technique applied in support vector machines and other machine learning techniques. Traditionally, the estimations of the market share attraction model are performed via a maximum likelihood estimation procedure under the assumption that the data are drawn from a normal distribution. The proposed method is shown to be a good candidate for forecasting method of the market share attraction model when normal distribution is not assumed. We apply the proposed method to forecast the market shares of 4 Korean car brands simultaneously and represent better performances than maximum likelihood estimation procedure.

Variance function estimation with LS-SVM for replicated data

심주용,박혜정,석경하 한국데이터정보과학회 2009 한국데이터정보과학회지 Vol.20 No.5

In this paper we propose a variance function estimation method for replicated data based on averages of squared residuals obtained from estimated mean function by the least squares support vector machine. Newton-Raphson method is used to obtain associated parameter vector for the variance function estimation. Furthermore, the cross validation functions are introduced to select the hyper-parameters which affect the performance of the proposed estimation method. Experimental results are then presented which illustrate the performance of the proposed procedure.

2P-661 겨울철 설해예방을 위한 근적외선 흡수 발열필름 연구

심주용,유국현,김정엽,김기혁 한국공업화학회 2017 한국공업화학회 연구논문 초록집 Vol.2017 No.0

Censored varying coefficient regression model using Buckley-James method

심주용,석경하 한국데이터정보과학회 2017 한국데이터정보과학회지 Vol.28 No.5

The censored regression using the pseudo-response variable proposed by Buckley and James has been one of the most well-known models. Recently, the varying coefficient regression model has received a great deal of attention as an important tool for modeling. In this paper we propose a censored varying coefficient regression model using Buckley-James method to consider situations where the regression coefficients of the model are not constant but change as the smoothing variables change. By using the formulation of least squares support vector machine (LS-SVM), the coefficient es- timators of the proposed model can be easily obtained from simple linear equations. Furthermore, a generalized cross validation function can be easily derived. In this paper, we evaluated the proposed method and demonstrated the adequacy through simulate data sets and real data sets.

Kernel Poisson regression for mixed input variables

심주용 한국데이터정보과학회 2012 한국데이터정보과학회지 Vol.23 No.6

An estimating procedure is introduced for kernel Poisson regression when the input variables consist of numerical and categorical variables, which is based on the penalized negative log-likelihood and the component-wise product of two different types of kernel functions. The proposed procedure provides the estimates of the mean function of the response variables, where the canonical parameter is linearly and/or nonlinearly related to the input variables. Experimental results are then presented which indicate the performance of the proposed kernel Poisson regression.