• ETRI Journal
• /
• 제27권6호
• /
• pp.747-758
• /
• 2005

The problem of selection among competing models has been a fundamental issue in statistical data analysis. Good fits to data can be misleading since they can result from properties of the model that have nothing to do with it being a close approximation to the source distribution of interest (for example, overfitting). In this study we focus on the preference among models from a family of polynomial regressors. Three decades of research has spawned a number of plausible techniques for the selection of models, namely, Akaike's Finite Prediction Error (FPE) and Information Criterion (AIC), Schwartz's criterion (SCH), Generalized Cross Validation (GCV), Wallace's Minimum Message Length (MML), Minimum Description Length (MDL), and Vapnik's Structural Risk Minimization (SRM). The fundamental similarity between all these principles is their attempt to define an appropriate balance between the complexity of models and their ability to explain the data. This paper presents an empirical study of the above principles in the context of model selection, where the models under consideration are univariate polynomials. The paper includes a detailed empirical evaluation of the model selection methods on six target functions, with varying sample sizes and added Gaussian noise. The results from the study appear to provide strong evidence in support of the MML- and SRM- based methods over the other standard approaches (FPE, AIC, SCH and GCV).

• 한국식생활문화학회지
• /
• 제25권5호
• /
• pp.516-524
• /
• 2010

Consumers are highly interested in processed fishery products that are healthy and superior in terms of convenience, nourishment, and taste. However, current domestic research on processed fishery products is marginal. We systematically analyzed consumer consumption patterns and the relationship to pursuit benefits, selection attributes, satisfaction levels, and reasons for purchase. Consumers considered product information the most important selection attribute, whereas convenience scored highest for pursuit benefits. Furthermore, the influences of selection attributes and pursuit benefits on satisfaction level and the reason for purchasing an item were analyzed using demographic properties as control variables. The variables that affected satisfaction level were residential district (region: B= -0.268, p<0.05.), recipe (B=0.098, p<0.05), nutrients (B=0.124, p<0.05), convenience (B=0.283, p<0.001), and economics (B=0.138, p<0.05). The variables affecting the reason for purchasing were nutrients (B=0.173, p<0.001), convenience (B=0.277, p<0.001) and satisfaction level (B=0.163, p<0.001). Pursuit intention had significant effects on purchase intention; however, selection attributes had no significant effect on purchase intention. Therefore, consumer satisfaction had a significant effect on purchase intention. This result showed that if consumers were satisfied, they intended to repurchase. Attempts to increase repurchases by consumer are needed by fulfilling consumer satisfaction. These data can be utilized as a fundamental reference for sales promotions.

• 한국컴퓨터정보학회논문지
• /
• 제13권1호
• /
• pp.1-10
• /
• 2008

이 기종의 네트워크와 시스템 자원으로 구성된 그리드 컴퓨팅 환경에서 대용량 데이터를 빠르고 정확하게 처리하기 위해서는 효과적인 그리드 자원선택이 필수적이다. 이를 위해 본 논문은 의사결정 트리 기법을 이용한 그리드 자원선택 시스템을 제안한다. 이 시스템은 자원 정보를 기록한 데이터 셋을 바탕으로 사용자들이 선택하는 자원들을 처리 할 데이터의 특성과 사용자의 요구사항으로 분석해서 자원선택을 위한 의사결정 트리를 구축한다. 그리드 사용자의 자원 요청 시 의사결정 트리를 탐색하여 사용자 요구 및 작업 특성에 적합한 자원들을 선택하여 작업을 할당함으로써 사용자 만족도를 향상시킴은 물론 전체 그리드 시스템의 성능을 개선한다. 실험결과는 본 논문에서 제안한 의사결정 트리 기반의 그리드 자원선택 시스템이 기존 그리드 자원선택 시스템인 Condor-G 및 Nimrod-G와 비교하여 더 높은 작업 처리율 및 자원 이용률과 더 적은 작업 손실 및 처리시간을 제공함으로써 그리드 자원선택 및 데이터 분산 처리에 효과적이라는 사실을 증명한다

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2005년도 춘계학술대회 논문집
• /
• pp.1774-1779
• /
• 2005

Most of materials receive force in using, therefore, the characteristics of materials must be considered in system design not to occur deformation or destruction. Mechanical properties about materials can be expressed as responsible level of material itself under the exterior operation. Main mechanical properties is strength, hardness, ductility and stiffness etc. Currently, among major measure facilities to measure such mechanical properties, advanced indentation technique has focused in industrial areas as reason of nondestructive and easy applications for mechanical tensile properties and evaluation of residual stress of materials. This study is to find the optimum experimental condition about residual stress advanced indentation technique for accurate analysis of the welded joint of steel materials through indentation load-depth curve obtained from cruciform specimen experiment. Optimum selection was applied to the welded joint of real steel materials to give non-equi-biaxial stress state and compared with general residual stress analyzing method for verification.

• 천문학회지
• /
• 제27권2호
• /
• pp.103-117
• /
• 1994

To extend the work of Gott, Park, and Lee (1989), statistical properties of gravitational lensing in a wide variety of cosmological models involving non-zero cosmological constant is investigated, using the redshifts of both lens and source and observed angular separation of images for gravitational lens systems. We assume singular isothermal sphere as lensing galaxy in homogenous and isotropic Friedmann­Lemaitre-Robertson- Walker universe, Schechter luminosity function, standard angular diameter distance formula and other galaxy parameters used in Fukugita and Turner (1991). To find the most adequate flat cosmological model and put a limit on the value of dimensionless cosmological constant $\lambda_0$, the mean value of the angular separation of images, probability distribution of angular separation and cumulative probability are calculated for given source and lens redshifts and compared with the observed values through several statistical methods. When there is no angular selection effect, models with highest value of $\lambda_0$ is preferred generally. When the angular selection effects are considered, the preferred model depends on the shape of the selection functions and statistical methods; yet, models with large $\lambda_0$ are preferred in general. However, the present data can not rule out any of the flat universe models with enough confidence. This approach can potentially select out best model. But at the moment, we need more data.

• 한국병원경영학회지
• /
• 제23권4호
• /
• pp.1-14
• /
• 2018

Purpose : Customers with loyalty are very important to hospitals for sustainable growth in their medical market. Individuals with loyalty are likely to visit same hospital repeatedly when they need medical services. This study was to identify factors associated with selection of specialty hospitals among customers with loyalty. Methods : The subjects of this study were 735 inpatients in 22 specialty hospitals in 6 designated fields(joints, spine, colorectal-anal, obstetrics and gynecology, ophthalmology, otolaryngology). Customer types classified as customers with high loyalty, neutral customers, and customers with low loyalty according to net promoter score(NPS). Factor analysis was conducted to classify 22 hospital selection factors into some similar properties. Logistic regression analysis was conducted to confirm the selection factors related to loyal customers. Findings : Most of specialty hospitals received high NPS of 8 points or higher in all the designated fields. Five factors associated with selection of specialty hospital are (1) hospital facilities and convenience, (2) trust in doctor and hospital, (3) rapidness of treatment, (4) hospital awareness, and (5) accessibility. As a result of logistic regression analysis, selection factors related to loyal customers were 'hospital facilities and convenience', 'trust in doctor and hospital' and 'rapidness of treatment'. Differences in the degree of importance of three selection factors by customer types appeared for each designated field. Practical Implications : This study confirms the high level of patient experience among inpatients of specialty hospitals. Factors associated with selection of hospital among inpatients with loyalty are 'facilities and convenience of hospitals', 'trust of doctor and hospital' and 'rapidness of treatment'. This study will be meaningful as basic data to systematically enhance the roles and functions of the health care system and to provide securing competitiveness according to designated fields in the management aspect of specialty hospitals.

• Journal of the Korean Statistical Society
• /
• 제22권2호
• /
• pp.361-368
• /
• 1993

In recent years, kernel type estimates are abundant. In this paper, we propose a bandwidth selection method for kernel regression of fixed design based on bootstrap procedure. Mathematical properties of proposed bootstrap-based bandwidth selection method are discussed. Performance of the proposed method for small sample case is compared with that of cross-validation method via a simulation study.

• Communications for Statistical Applications and Methods
• /
• 제21권1호
• /
• pp.61-68
• /
• 2014

We propose binary classification methods by modifying logistic regression classification. We use variable selection procedures instead of original variables to select the principal components. We describe the resulting classifiers and discuss their properties. The performance of our proposals are illustrated numerically and compared with other existing classification methods using synthetic and real datasets.

• 대한수학회보
• /
• 제36권4호
• /
• pp.723-736
• /
• 1999

The paper is devoted to generalizations of continuity of set-valued mappings and some properties of hypertopologies on the collection of some subsets of a topological space. It is also dedicated to continuous selection theorems without relatively higher separation axioms. More precisely, we give characterizations of $\lambda$-collectionwise normality using continuous functions as in Michael's papers.

• 한국경영과학회지
• /
• 제6권2호
• /
• pp.35-49
• /
• 1981

This paper is concerned with developing a Multi-period Behavioral Model for the portfolio selection problem. The unique feature of the model is that it treats a number of factors and decision variables considered germane in decision making on an interrelated basis. The formulated problem has the structure of a Chance Constrained programming Model. Then empoloying arguments of Central Limit Theorem and normality assumption the stochastic model is reduced to that of a Non-Linear Programming Model. Finally, a number of interesting properties for the reduced model are established.