• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2015년도 학술발표회
• /
• pp.606-609
• /
• 2015

본 연구에서는 support vector regression (SVR) 및 매개변수 최적화 알고리즘을 이용한 하천수위 예측모델을 구축하고 이를 실제 유역에 적용하여 모델 효율성을 평가하였다. 여기서, SVR은 하천수위를 예측하기 위한 예측모델로서 채택되었으며, 커널함수 (Kernel function)로서는 radial basis function (RBF)을 선택하였다. 최적화 알고리즘은 SVR의 최적 매개변수 (C?, cost parameter or regularization parameter; ${\gamma}$, RBF parameter; ${\epsilon}$, insensitive loss function parameter)를 탐색하기 위하여 적용되었다. 매개변수 최적화 알고리즘으로는 grid search (GS), genetic algorithm (GA), particle swarm optimization (PSO), artificial bee colony (ABC) 알고리즘을 채택하였으며, 비교분석을 통해 최적화 알고리즘의 적용성을 평가하였다. 또한 SVR과 최적화 알고리즘을 결합한 모델 (SVR-GS, SVR-GA, SVR-PSO, SVR-ABC)은 기존에 수자원 분야에서 널리 적용되어온 신경망(Artificial neural network, ANN) 및 뉴로퍼지 (Adaptive neuro-fuzzy inference system, ANFIS) 모델과 비교하였다. 그 결과, 모델 효율성 측면에서 SVR-GS, SVR-GA, SVR-PSO 및 SVR-ABC는 ANN보다 우수한 결과를 나타내었으며, ANFIS와는 비슷한 결과를 나타내었다. 또한 SVR-GA, SVR-PSO 및 SVR-ABC는 SVR-GS보다 상대적으로 우수한 결과를 나타내었으며, 모델 효율성 측면에서 SVR-PSO 및 SVR-ABC는 가장 우수한 모델 성능을 나타내었다. 따라서 본 연구에서 적용한 매개변수 최적화 알고리즘은 SVR의 매개변수를 최적화하는데 효과적임을 확인할 수 있었다. SVR과 최적화 알고리즘을 이용한 하천수위 예측모델은 기존의 ANN 및 ANFIS 모델과 더불어 하천수위 예측을 위한 효과적인 도구로 사용될 수 있을 것으로 판단된다.

• Journal of the Korean Data and Information Science Society
• /
• 제9권2호
• /
• pp.357-363
• /
• 1998

In this paper we propose a nonparametric lack of fit test based on the bootstrap method for testing the null parametric linear model by using linear smoothers. Most of existing nonparametric test statistics are based on the residuals. Our test is based on the centered bootstrap residuals. Power performance of proposed bootstrap lack of fit test is investigated via Monte carlo simulation.

• Journal of the Korean Statistical Society
• /
• 제31권1호
• /
• pp.109-120
• /
• 2002

In this paper, we show the weak convergence of U-empirical processes for two sample problem. We use the result to show the asymptotic normality for the generalized dodges-Lehmann estimates with the Bahadur representation for quantifies of U-empirical distributions. Also we consider the asymptotic normality for the test statistics in a simple way.

• Journal of the Korean Data and Information Science Society
• /
• 제21권2호
• /
• pp.363-369
• /
• 2010

In this paper we propose a mixed-effects least squares support vector regression (LS-SVR) for longitudinal data. We add a random-effect term in the optimization function of LS-SVR to take random effects into LS-SVR for analyzing longitudinal data. We also present the model selection method that employs generalized cross validation function for choosing the hyper-parameters which affect the performance of the mixed-effects LS-SVR. A simulated example is provided to indicate the usefulness of mixed-effect method for analyzing longitudinal data.

• Journal of applied mathematics & informatics
• /
• 제9권1호
• /
• pp.261-275
• /
• 2002

Besides asymptotic method, the method of orthogonal polynomials has been used to obtain the solution of the Fredholm integral equation. The principal (singular) part of the kerne1 which corresponds to the selected domain of parameter variation is isolated. The unknown and known functions are expanded in a Chebyshev polynomial and an infinite a1gebraic system is obtained.

• Communications for Statistical Applications and Methods
• /
• 제16권6호
• /
• pp.989-995
• /
• 2009

Basu et al. (1998) proposed the minimum divergence estimating method which is free from using the painful kernel density estimator. Their proposed class of density power divergences is indexed by a single parameter $\alpha$ which controls the trade-off between robustness and efficiency. In this article, (1) we introduce a new large class the minimum squared distance which includes from the minimum Hellinger distance to the minimum $L_2$ distance. We also show that under certain conditions both the minimum density power divergence estimator(MDPDE) and the minimum squared distance estimator(MSDE) are asymptotically equivalent and (2) in finite samples the MDPDE performs better than the MSDE in general but there are some cases where the MSDE performs better than the MDPDE when estimating a location parameter or a proportion of mixed distributions.

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 2013년도 추계학술대회 논문집
• /
• pp.331-337
• /
• 2013

Acoustic Emission technique is widely applied to develop the early fault detection system, and the problem about a signal processing method for AE signal is mainly focused on. In the signal processing method, envelope analysis is a useful method to evaluate the bearing problems and Wavelet transform is a powerful method to detect faults occurred on rotating machinery. However, exact method for AE signal is not developed yet. Therefore, in this paper two methods which are Hilbert transform and DET for feature extraction. In addition, we evaluate the classification performance with varying the parameter from 2 to 15 for feature selection DET, 0.01 to 1.0 for the RBF kernel function of SVR, and the proposed algorithm achieved 94% classification accuracy with the parameter of the RBF 0.08, 12 feature selection.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제2권1호
• /
• pp.1-6
• /
• 2013

소프트웨어의 결함 경향 모듈 예측을 위해 SVM 분류기가 우수한 성능을 보인다는 연구들이 많지만, SVM에서 필요한 파라미터 선정 시 매 커널마다 다르게 선정해야 하고, 파라미터의 변경에 따른 결과예측을 위해 알고리즘을 반복적으로 수행해야 하는 불편함이 있다. 따라서 본 논문에서는 SVM의 파라미터 선정 시 유전알고리즘을 이용하여 스스로 찾게 하는 GA-SVM 모델을 구현하였다. 그리고 분류 성능 비교를 위해 신경망의 역전파알고리즘을 이용하여 분류했던 기존 논문과 비교 분석한 결과, GA-SVM 모델의 성능이 더 우수함을 확인하였다.

• International Journal of Advanced Culture Technology
• /
• 제10권1호
• /
• pp.236-241
• /
• 2022

We consider methods of estimating a binary regression function using a nonparametric kernel estimation when there is only one covariate. For this, the Nadaraya-Watson estimation method using single and double bandwidths are used. For choosing a proper smoothing amount, the cross-validation and plug-in methods are compared. In the real data analysis for case study, German credit data and heart disease data are used. We examine whether the nonparametric estimation for binary regression function is successful with the smoothing parameter using the above two approaches, and the performance is compared.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권1호
• /
• pp.413-435
• /
• 2018

Although the accuracy of handwritten character recognition based on deep networks has been shown to be superior to that of the traditional method, the use of an overly deep network significantly increases time consumption during parameter training. For this reason, this paper took the training time and recognition accuracy into consideration and proposed a novel handwritten character recognition algorithm with newly designed network structure, which is based on an extended nonlinear kernel residual network. This network is a non-extremely deep network, and its main design is as follows:(1) Design of an unsupervised apriori algorithm for intra-class clustering, making the subsequent network training more pertinent; (2) presentation of an intermediate convolution model with a pre-processed width level of 2;(3) presentation of a composite residual structure that designs a multi-level quick link; and (4) addition of a Dropout layer after the parameter optimization. The algorithm shows superior results on MNIST and SVHN dataset, which are two character benchmark recognition datasets, and achieves better recognition accuracy and higher recognition efficiency than other deep structures with the same number of layers.