• Journal of Korean Institute of Industrial Engineers
• /
• v.32 no.2
• /
• pp.74-81
• /
• 2006

In this paper, we deal with the separation of data by concurrently determined, piecewise nonlinear discriminant functions. Toward the end, we develop a new $l_1$-distance norm error metric and cast the problem as a mixed 0-1 integer and linear programming (MILP) model. Given a finite number of discriminant functions as an input, the proposed model considers the synergy as well as the individual role of the functions involved and implements a simplest nonlinear decision surface that best separates the data on hand. Hence, exploiting powerful MILP solvers, the model efficiently analyzes any given data set for its piecewise nonlinear separability. The classification of four sets of artificial data demonstrates the aforementioned strength of the proposed model. Classification results on five machine learning benchmark databases prove that the data separation via the proposed MILP model is an effective supervised learning methodology that compares quite favorably to well-established learning methodologies.

• Communications for Statistical Applications and Methods
• /
• v.7 no.1
• /
• pp.225-232
• /
• 2000

The linear and quadratic discrimination functions based on normal theory are widely used to classify an observation to one of predefined groups. But the discriminant functions are sensitive to outliers. A high breakdown procedure to estimate location and scatter of multivariate data is the minimum volume ellipsoid or MVE estimator To obtain high breakdown classifiers outliers in multivariate data are detected by using the robust Mahalanobis distance based on MVE estimators and the weighted estimators are inserted in the functions for classification. A samll-sample MOnte Carlo study shows that the high breakdown robust procedures perform better than the classical classifiers.

• Journal of Ecology and Environment
• /
• v.36 no.3
• /
• pp.205-210
• /
• 2013

Cicuta virosa L. (Apiaceae) is a perennial emergent plant designated as an endangered species in South Korea. According to the former records, only four natural habitats remain in South Korea. A former study suggested that three of four populations (Pyeongchang: PC, Hoengseong: HS, Gunsan: GS) would be classified as different ecotypes based on their different morphological characteristics and life cycle under different environmental conditions. To evaluate this suggestion, we estimated genetic diversity in each population and distance among three populations by random amplification of polymorphic DNA. Seven random primers generated a total of 61 different banding positions, 36 (59%) of them were polymorphic. Nei's gene diversity and the Shannon diversity index increased in the order of PC < HS < GS, which is the same order of population size. In the two-dimensional (2D) plot of first two principal components in principal component analysis with the presence of 61 loci, individuals could be grouped as three populations easily (proportion of variance = 0.6125). Nei's genetic distance for the three populations showed the same tendency with the geographical distance within three populations. And it is also similar to the result of discriminant analysis with the morphological or life-cycle factors from the previous study. From the results, we concluded that three different populations of C. virosa should be classified as ecotypes based on not only morphology and phenology but genetic differences in terms of diversity and distance as well.

• Korean Journal of Human Ecology
• /
• v.14 no.3
• /
• pp.467-473
• /
• 2005

The integrative Internet program was developed for the selection of optimal brassiere components according to the breast types of middle-aged women. In this program, a customer is classified into a certain group through three steps of a discriminant analysis. Three variables used in the analysis include breast volume, radii of curvature of under-breast curve, and a distance between inner breast points. Using individual data of three variables, the optimal brassiere components, i.e. brassiere cup size, curvature of front panel and wire, distance between cups, are suggested for each customer. Discrimination of breast types using only 2D measurements is also included for those who do not have easy access to a 3D measurement device.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.23 no.4
• /
• pp.737-742
• /
• 2013

The surveillance system has been developed to be intelligent which can judge and cope by itself using human recognition technique. The existing face recognition is excellent at a short distance but recognition rate is reduced at a long distance. In this paper, we analyze the performance of face recognition according to interpolation and face recognition algorithm in face recognition using the multiple distance face images to training. we use the nearest neighbor, bilinear, bicubic, Lanczos3 interpolations to interpolate face image and PCA and LDA to face recognition. The experimental results show that LDA-based face recognition with bilinear interpolation provides performance in face recognition.

• Journal of the Korean Society of Safety
• /
• v.31 no.1
• /
• pp.95-104
• /
• 2016

In general, the crash reduction effect of the rumble strip is reported to be about 30% in Korea, while it is about 40-60% in the United States. However, the effect is erroneously overestimated because the simple comparison was only made before and after the installation. Accordingly, this study will reassess the crash reduction effect of the rumble strip. The study will also examine the former's geometric characteristics as well as its effect on the causes of the crash. This study analyzed the crash effect while taking into consideration the changes in the horizontal and vertical alignment, including the width of pavement shoulders, using the crash data for two years before and after the installation of the rumble strip. The types of crash caused by the rumble strip were identified using the classification discriminant function. The crash effect on the rumble strip is estimated to be 28.3%, but the pure effect, with the exception of the effect by other elements, was analyzed to be 7.4%. For each expressway design element, the downhill section (2.0-3.0%), the section with less than 3,000 m and more than 10,000 m of the curve radius, and the section with less than 3.0 m of the pavement shoulder width were found to be effective in crash reduction. For each cause of crash, the rumble strip was analyzed to be effective in the reduction of crash caused by "not keeping the safe distance", "sleeping", "negligence in keeping eyes forward", and "excessive handle operation". In particular, the rumble strip was analyzed and seen to be especially effective in preventing crash caused by "not keeping a safe distance," and "sleeping". The installation of the rumble strip was found to be effective in the prevention of crash caused by "not keeping the safe distance" and "sleeping". The results of this study may thus be used in deciding the causes of crash and the installation location of the tailored rumble strip that would be suitable for the geometric characteristics of the roads. This study will also be helpful in the establishment of future traffic safety measures.

• Economic and Environmental Geology
• /
• v.36 no.3
• /
• pp.201-212
• /
• 2003

Recently, the ability of earthquake detection in the Kyungsang Basin of southeastern Korean Peninsula is greatly improved since seismic stations including seismic network of KIGAM(Korea Institute of Geoscience and Mineral Resources) have been significantly increased. However, a large number of signals from explosions are recorded because of frequent medium to large chemical explosions. The discrimination between natural earthquakes and explosions in the Basin has become an important issue. High frequency local records from 43 earthquakes and 43 explosions with comparable magnitude are selected to establish a reliable discrimination technique in the Basin. Several discrimination techniques in spectral domain using spectral amplitude ratios among Pg, Sg, and Lg waves are widely examined with tile selected data. Among them the Pg/Lg spectral ratio method is appeared to be a good discrimination technique to improve the discrimination power. Multivariate discriminant analysis is also applied to the Pg/Lg spectral ratios. The discrimination power of the Pg/Lg ratios for distance corrected three component record compared to uncorrected vertical component one shows distinct improvement. In the frequency band 4 to 14 Hz, Pg/Lg spectral ratio for distance corrected three component record provides discrimination power with a total misclassification probability of only 0.89%.

• Journal of Korea Multimedia Society
• /
• v.10 no.7
• /
• pp.923-933
• /
• 2007

In this paper, we present a method for performance enhancement of the marker detection system by using SVM(Support Vector Machine) and LDA(Linear Discriminant Analysis). It converts the input image to a binary image and extracts contours of objects in the binary image. After that, it approximates the contours to a list of line segments. It finds quadrangle by using geometrical features which are extracted from the approximated line segments. It normalizes the shape of extracted quadrangle into exact squares by using the warping technique and scale transformation. It extracts feature vectors from the square image by using principal component analysis. It then checks if the square image is a marker image or a non-marker image by using a SVM classifier. After that, it computes feature vectors by using LDA for the extracted marker images. And it calculates the distance between feature vector of input marker image and those of standard markers. Finally, it recognizes the marker by using minimum distance method. Experimental results show that the proposed method achieves enhancement of recognition rate with smaller feature vectors by using LDA and it can decrease false detection errors by using SVM.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.1
• /
• pp.187-196
• /
• 2015

The team's performances were analyzed by evaluating the scores gained by their offense and the scores allowed by their defense. To evaluate the team's attacking and defending abilities, we also considered the factors that contributed the team's gained points or the opposing team's gained points? In order to analyze the outcome of the games, three prediction models were used such as decision trees, logistic regression, and discriminant analysis. As a result, the factors associated with the defense showed a decisive influence in determining the game results. We analyzed the offense and defense by using the response variable. This showed that the major factors predicting the offense were non-stop pass and attack speed and the major factor predicting the defense were the distance between right and left players and the distance between front line attackers and rearmost defenders during the game.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.20 no.2
• /
• pp.282-297
• /
• 1996

The effective construction for ready-made clothes is one of the central concerns of both consumers and manufactuers in today's apparel industry. In order to reduce the burden of stocks and increase clothing fitness, systematic information on typical body sizes and somatotypes is essential. The purpose of this study i-: to provide basic data on young women's somatotypes for form designers and pattern makers. The subjects of the survey were 310 women of 18 to 26 years old. The study collected 84 anthropometric data for each Person. The data was analyzed by using of the multivariate method. The factor analysis was utilized in regard to the 65 items obtained from anthropometric measurement respectively. The principal component analysis was applied to the data with orthogonal rotation after extraction. The factor scores used in the factor analysis became the basis of determining the value of each variable of the cluster analysis. The cluster analysis was applied for identifying typical somatotypes. Ward's minimum variance method was applied for the purpose of extracting distance metrix by the standardized Euclidean distance. The element forming each cluster can be subdivided into several sets by crosstabulation which is obtained by the fastclus of the SAS. This research has demonstrated 3 distinctive types of silhouette contour of the trunk. Incidentally it also identified 4 of the lower body from the waistline to thigh contour respectively. The discriminant analysis showed that the most significant discriminant factor of the trunk classification were side neck point -1 scapular -1 waistiline length and waist girth. In Korea, the average somatotype of female college students tends to be tall, slim and straight. Reviewing the relationship between the classifications of three parts of body, they are related to each other to some extent but their distribution are not constant. Therefore, in view of clothing construction, a proper separation of the body surface is a necessity.