• Communications for Statistical Applications and Methods
• /
• v.28 no.2
• /
• pp.189-204
• /
• 2021

In a regression analysis, a single best model is usually selected among several candidate models. However, it is often useful to combine several candidate models to achieve better performance, especially, in the prediction viewpoint. Model combining methods such as stacking and Bayesian model averaging (BMA) have been suggested from the perspective of averaging candidate models. When the candidate models include a true model, it is expected that BMA generally gives better performance than stacking. On the other hand, when candidate models do not include the true model, it is known that stacking outperforms BMA. Since stacking and BMA approaches have different properties, it is difficult to determine which method is more appropriate under other situations. In particular, it is not easy to find research papers that compare stacking and BMA when regression model assumptions are violated. Therefore, in the paper, we compare the performance among model averaging methods as well as a single best model in the linear regression analysis when standard linear regression assumptions are violated. Simulations were conducted to compare model averaging methods with the linear regression when data include outliers and data do not include them. We also compared them when data include errors from a non-normal distribution. The model averaging methods were applied to the water pollution data, which have a strong multicollinearity among variables. Simulation studies showed that the stacking method tends to give better performance than BMA or standard linear regression analysis (including the stepwise selection method) in the sense of risks (see (3.1)) or prediction error (see (3.2)) when typical linear regression assumptions are violated.

• Journal of the Society of Naval Architects of Korea
• /
• v.42 no.5 s.143
• /
• pp.534-541
• /
• 2005

Until now, Korean shipyards have accumulated a great amount of data. But they do not have appropriate tools to utilize the data in practical works. Engineering data contains experts' experience and know-how in its own. It is very useful to extract knowledge or information from the accumulated existing data by using data mining technique This paper treats an evolutionary computation based on genetic programming (GP), which can be one of the components to realize data mining. The paper deals with linear models of GP for the regression or approximation problem when given learning samples are not sufficient. The linear model, which is a function of unknown parameters, is built through extracting all possible base functions from the standard GP tree by utilizing the symbolic processing algorithm. In addition to a standard linear model consisting of mathematic functions, one variant form of a linear model, which can be built using low order Taylor series and can be converted into the standard form of a polynomial, is considered in this paper. The suggested model can be utilized as a designing tool to predict design parameters with small accumulated data.

• Journal of Welding and Joining
• /
• v.17 no.4
• /
• pp.85-92
• /
• 1999

In GMAW, the spatter is generated because of the variation of the arc state. If the arc state is quantitatively assessed, the control method to make the spatter be reduced is able to develop. This study was attempted to develop the optimal model that could estimate the arc state quantitatively. To do this, the generated spatters was captured under the limited welding conditions, and the waveforms of the arc voltage and of the welding current were collected. From the collected waveforms, the waveform factors and their standard deviations were produced, and the linear and non-linear regression models constituted using the factors and their standard deviations are proposed to estimate the arc state. the performance test to the proposed models was practiced. Obtained results are as follow. From the results of correlation analysis between the factors and the amount of the generated spatters, the standard deviations of the waveform factors have more the multiple regression coefficients than the waveform factors. Because the correlation coefficient between T and {TEX}$T_{a}${/TEX}, and s[T] and s[{TEX}$T_{a}${/TEX}] was nearly one, it was found that these factors have the same effect to the spatter generation. In the regression models to estimate the arc state, it was fond that the linear and the non linear models were also consisted of similar factors. In addition, the linear regression model was assessed the optimal model for estimating the arc state because the variance of data was narrow and multiple regression coefficient was highest among the models. But in the welding conditions which the amount of the generated spatters were small, it was found that the non linear regression model had better the estimation performance for the spatter generation than the linear.

• The Korean Journal of Applied Statistics
• /
• v.33 no.6
• /
• pp.813-821
• /
• 2020

Recently, research using unstructured data such as images and videos has been actively conducted in various fields. Edge detection is one of the most useful image enhancement techniques to improve the quality of the image process. However, it is very difficult to perform edge detection in noise images because the edges and noise having high frequency components. This paper uses a linear model and standard deviation as an effective edge detection method for noise images. The edge is detected by the difference between the standard deviation of the pixels included in the pixel block and the standard deviation of the residual obtained by fitting the linear model. The results of edge detection are compared with the results of the Sobel edge detector. In the original image, the Sobel edge detection result and the proposed edge detection result are similar. Proposed method was confirmed that the edge with reduced noise was detected in the various levels of noise images.

• Proceedings of the KIEE Conference
• /
• 2006.07d
• /
• pp.1854-1855
• /
• 2006

In this paper, the Takagi-Sugeno (TS) fuzzy state estimation scheme, which is suggested for a steady state estimator using standard Kalman filter theory with uncertainties. In that case, the steady state with uncertain can be represented by the TS fuzzy model structure, which is further rearranged to give a set of uncertain linear model using standard Kalman filter theory. And then the unknown uncertainty is regarded as an additive process noise. To optimize fuzzy system, we utilize the genetic algorithm. The steady state solutions can be found for proposed linear model then the linear combination is used to derive a global model. The proposed state estimator is demonstrated on a truck-trailer.

• Journal of the Korean Data and Information Science Society
• /
• v.15 no.1
• /
• pp.129-139
• /
• 2004

We consider the problem of testing for outliers in generalized linear model. We proceed by first specifying a mean shift outlier model, assuming the suspect set of ourliers is known. Given this model, we discuss standard approaches to obtaining score test for outliers as an alternative to the likelihood ratio test.

• International Journal of Control, Automation, and Systems
• /
• v.6 no.1
• /
• pp.24-34
• /
• 2008

A novel neural network model, termed the standard neural network model (SNNM), similar to the nominal model in linear robust control theory, is suggested to facilitate the synthesis of controllers for delayed (or non-delayed) nonlinear systems composed of neural networks. The model is composed of a linear dynamic system and a bounded static delayed (or non-delayed) nonlinear operator. Based on the global asymptotic stability analysis of SNNMs, Static state-feedback controller and dynamic output feedback controller are designed for the SNNMs to stabilize the closed-loop systems, respectively. The control design equations are shown to be a set of linear matrix inequalities (LMIs) which can be easily solved by various convex optimization algorithms to determine the control signals. Most neural-network-based nonlinear systems with time delays or without time delays can be transformed into the SNNMs for controller synthesis in a unified way. Two application examples are given where the SNNMs are employed to synthesize the feedback stabilizing controllers for an SISO nonlinear system modeled by the neural network, and for a chaotic neural network, respectively. Through these examples, it is demonstrated that the SNNM not only makes controller synthesis of neural-network-based systems much easier, but also provides a new approach to the synthesis of the controllers for the other type of nonlinear systems.

• Korean Management Science Review
• /
• v.28 no.1
• /
• pp.141-158
• /
• 2011

This study concerns with an applicability of the management science approach to the score adjustment among the College Scholastic Aptitude Test(CSAT) optional subjects. A linear programming model is developed to minimize the sum of score distortions between optional subjects. Based on the analysis of the 377,089 CSAT(2010) applicants' performances in social science test section, this study proposes a new approach for the score equating or linking method of the educational measurement theory. This study makes up for the weak points in the previous linear programming model. First, the model utilize the standard score which we can get. Second, the model includes a goal programming concept which minimizes the gap between the adjusting goal and the result of the adjustment. Third, the objective function of the linear programing is the weighted sum of the score distortion and the number of applicants. Fourth, the model is applied to the score adjustment problem for the whole 11 optional subjects of the social science test section. The suggested linear programming model is a generalization of the multi-tests linking problem. So, the approach is consistent with the measurement theory for the two tests and can be applied to the optional three or more tests which do not have a common anchor test or a common anchor group. The college admission decision with CSAT score can be improved by using the suggested linear programming model.

• Communications for Statistical Applications and Methods
• /
• v.10 no.1
• /
• pp.125-134
• /
• 2003

Standard estimation methods for linear mixed models are sensitive to influential observations. However, tools and concepts for linear mixed model diagnostics are rudimentary until now and research is heavily demanded in linear mixed models. In this paper, we consider two diagnostics to evaluate the effects of individual observations in the estimation of fixed effects for linear mixed models. Those are Cook's distance and COVRATIO. Results of our limited simulation study suggest that the Cook's distance is not good statistical quantity in linear mixed models. Also calibration point for COVRATIO seems to be quite conservative.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.58 no.7
• /
• pp.1411-1417
• /
• 2009

In this paper, we consider the design of $H_{\infty}$ control of linear discrete-time systems having no mathematical model. The basic approach is to use Q-learning which is a reinforcement learning method based on actor-critic structure. The model-free control design is to use not the mathematical model of the system but the informations on states and inputs. As a result, the derived iterative algorithm is expressed as linear matrix inequalities(LMI) of measured data from system states and inputs. It is shown that, for a sufficiently rich enough disturbance, this algorithm converges to the standard $H_{\infty}$ control solution obtained using the exact system model. A simple numerical example is given to show the usefulness of our result on practical application.