• The KIPS Transactions:PartD
• /
• v.15D no.6
• /
• pp.767-776
• /
• 2008

Microarray technology is extensively being used in experimental molecular biology field. Microarray experiments generate quantitative expression measurements for thousands of genes simultaneously, which is useful for the phenotype classification of many diseases. One of the two major problems in microarray data classification is that the number of genes exceeds the number of tissue samples. The other problem is that current methods generate classifiers that are accurate but difficult to interpret. Our paper addresses these two problems. We performed a direct integration of individual microarrays with same biological objectives by transforming an expression value into a rank value within a sample and generated rank-comparison decision rules with variable number of genes for cancer classification. Our classifier is an ensemble method which has k top scoring decision rules. Each rule contains a number of genes, a relationship among involved genes, and a class label. Current classifiers which are also ensemble methods consist of k top scoring decision rules. However these classifiers fix the number of genes in each rule as a pair or a triple. In this paper we generalized the number of genes involved in each rule. The number of genes in each rule is in the range of 2 to N respectively. Generalizing the number of genes increases the robustness and the reliability of the classifier for the class prediction of an independent sample. Also our classifier is readily interpretable, accurate with small number of genes, and shed a possibility of the use in a clinical setting.

• Journal of Korean Academy of Nursing Administration
• /
• v.5 no.3
• /
• pp.513-524
• /
• 1999

The shortage of nursing personnel was become one of the most serious problems in operating pediatric oncology nursing unit which was the first pediatric oncology nursing unit in Korea. The purpose of this study was to estimate the optimal number of nursing personnel by calculating nursing care hours. The subjects were 13 staff nurses and inpatients of pediatric oncology nursing unit at Seoul National University Hospital during the period of May 20, 1996, to June 2, 1996. The number of nurses' duty was 132, the number of patients treated was 1288 for these 2 weeks. The tools used for this study were pediatric patient classification indexes and direct & indirect care indexes. Each nurse measured the time that they spent for their activities by self record under the supervision of their nurse manager. The method used to calculate the number of nursing personnel was multiplication of the average number of nursing care hours per patient per day with the number of patients. Percentage, average, t-test, F-test were used for data analysis. The results of this study were as follows : 1) The distribution of patient class : Class I & II none, Class III 86.8%. Class IV 12.9% 2) Direct nursing care hours for a patient per shift according to patient classification: Class III : 27.64 minutes, Class IV : 54.64 minutes The average direct nursing service hours for a patient per shift(3 shift) was 31.54 minutes(94.62 m/day). The average indirect nursing service hours for each patient per duty(3 shift) is 21.3 minutes (63. 91 m/day). 3) The average nursing hours for a patient per duty was 52.80 minutes(2.64h/day). 4) The group of administering medications in direct care activities showed the highest percentage (38.9%). Checking vital signs among observation took the most time am.ong each direct care activity (6.88 minutes for a patient per duty). 5) Charting took the most time of each indirect care activity(52.53 minutes/ duty/nurse). 6) The average personal time per duty is 29.40 minutes, which 'was below 30 minutes of this hospital regulations. 7) The average nursing hours that a nurse provided for a duty was 8.60 hours, which meant that a nurse worked 1.10 hours overtime. 8) Standardizing to a 33 bed to a unit, 17 nurses were needed at the present nursing level.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.7
• /
• pp.299-306
• /
• 2022

In recent, sensors embedded in robots, equipment, and circuits have become common, and research for diagnosing device failures by learning measured sensor data is being actively conducted. This failure diagnosis study is divided into a classification model for predicting failure situations or types and a regression model for numerically predicting failure conditions. In the case of a classification model, it simply checks the presence or absence of a failure or defect (Class), whereas a regression model has a higher learning difficulty because it has to predict one value among countless numbers. So, the reason that regression modeling is more difficult is that there are many irregular situations in which it is difficult to determine one output from a similar input when predicting by matching input and output. Therefore, in this paper, we focus on input and output data with periodicity, analyze the input/output relationship, and secure regularity between input and output data by performing sliding window-based input data patterning. In order to apply the proposed method, in this study, current and temperature data with periodicity were collected from MMC(Modular Multilevel Converter) circuit system and learning was carried out using ANN. As a result of the experiment, it was confirmed that when a window of 2% or more of one cycle was applied, performance of 97% or more of fit could be secured.

• Korean Journal of Remote Sensing
• /
• v.20 no.6
• /
• pp.383-396
• /
• 2004

The application of spatial statistics to obtain the spatial uncertainty distributions in classification of remote sensing images is investigated in this paper. Two quantitative methods are presented for describing two kinds of uncertainty; one related to class assignment and the other related to the connection of reference samples. Three quantitative indices are addressed for the first category of uncertainty. Geostatistical simulation is applied both to integrate the exhaustive classification results with the sparse reference samples and to obtain the spatial uncertainty or accuracy distributions connected to those reference samples. To illustrate the proposed methods and to discuss the operational issues, the experiment was done on a multi-sensor remote sensing data set for supervised land-cover classification. As an experimental result, the two quantitative methods presented in this paper could provide additional information for interpreting and evaluating the classification results and more experiments should be carried out for verifying the presented methods.

• Proceedings of the IEEK Conference
• /
• 2001.06d
• /
• pp.231-234
• /
• 2001

This paper proposes a face recognition technique that effectively combines elastic graph matching (EGM) and Fisherface algorithm. EGM as one of dynamic lint architecture uses not only face-shape but also the gray information of image, and Fisherface algorithm as a class specific method is robust about variations such as lighting direction and facial expression. In the proposed face recognition adopting the above two methods, the linear projection per node of an image graph reduces dimensionality of labeled graph vector and provides a feature space to be used effectively for the classification. In comparison with a conventional method, the proposed approach could obtain satisfactory results in the perspectives of recognition rates and speeds. Especially, we could get maximum recognition rate of 99.3% by leaving-one-out method for the experiments with the Yale Face Databases.

• 한국HCI학회:학술대회논문집
• /
• 2008.02a
• /
• pp.504-509
• /
• 2008

Support Vector Machine (SVM) is one of powerful learning machine and has been applied to varying task with generally acceptable performance. The success of SVM for classification tasks in one domain is affected by features which represent the instance of specific class. Given the representative and discriminative features, SVM learning will give good generalization and consequently we can obtain good classifier. In this paper, we will assess the problem of feature choices for human detection tasks and measure the performance of each feature. Here we will consider HOG-family feature. As a natural extension of SVM, we combine SVM with Principal Component Analysis (PCA) to reduce dimension of features while retaining most of discriminative feature vectors.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2005.09a
• /
• pp.23-28
• /
• 2005

The high-dimensionality and insufficiency of gene expression profiles and proteomic profiles makes feature selection become a critical step in efficiently building accurate models for cancer problems based on such data sets. In this paper, we use a method, called Discrete Function Learning algorithm, to find discriminatory feature vectors based on information theory. The target feature vectors contain all or most information (in terms of entropy) of the class attribute. Two data sets are selected to validate our approach, one leukemia subtype gene expression data set and one ovarian cancer proteomic data set. The experimental results show that the our method generalizes well when applied to these insufficient and high-dimensional data sets. Furthermore, the obtained classifiers are highly understandable and accurate.

• Korean Journal of Remote Sensing
• /
• v.21 no.6
• /
• pp.457-468
• /
• 2005

Increasing of impervious surface resulting from urban development has negative impacts on urban environment. Therefore, it is absolutely necessary to estimate and quantify the temporal and spatial aspects of impervious area for study of urban environment. In many cases, conventional image classification methods have been used for analysis of impervious surface fraction. However, the conventional classification methods have shortcoming in estimating impervious surface. The DN value of the each pixel in imagery is mixed result of spectral character of various objects which exist in surface. But conventional image classification methods force each pixel to be allocated only one class. And also after land cover classification, it is requisite to additional work of calculating impervious percentage value in each class item. This study used the spectral mixture analysis to overcome this weakness of the conventional classification methods. Four endmembers, vegetation, soil, low albedo and high albedo were selected to compose pure land cover objects. Impervious surface fraction was estimated by adding low albedo and high albedo. The study area is the Tanchon watershed which has been rapidly changed by the intensive development of housing. Landsat imagery from 1988, 1994 to 2001 was used to estimate impervious surface fraction. The results of this study show that impervious surface fraction increased from $15.6\%$ in 1988, $20.1\%$ in 1994 to $24\%$ in 2001. Results indicate that impervious surface fraction can be estimated by spectral mixture analysis with promising accuracy.

• Korean Journal of Remote Sensing
• /
• v.32 no.3
• /
• pp.235-244
• /
• 2016

In large-area crop classification with MODIS data, a mixed pixel problem caused by the low resolution of MODIS data has been one of main issues. To mitigate this problem, this paper proposes a hierarchical classification algorithm that selectively classifies the specific crop class of interest by using their spectral characteristics. This selective classification algorithm can reduce mixed pixel effects between crops and improve classification performance. The methodological developments are illustrated via a case study in Jilin city, China with MODIS Normalized Difference Vegetation Index (NDVI) and Near InfRared (NIR) reflectance datasets. First, paddy fields were extracted from unsupervised classification of NIR reflectance. Non-paddy areas were then classified into corn and bean using time-series NDVI datasets. In the case study result, the proposed classification algorithm showed the best classification performance by selectively classifying crops having similar spectral characteristics, compared with traditional direct supervised classification of time-series NDVI and NIR datasets. Thus, it is expected that the proposed selective hierarchical classification algorithm would be effectively used for producing reliable crop maps.

• Journal of Korean Neurosurgical Society
• /
• v.40 no.5
• /
• pp.323-329
• /
• 2006

Objective : This study was designed to analyze surgical strategies for patients with intractable supplementary sensorimotor area[SSMA] seizures. Methods : Seventeen patients who had surgical treatment were reviewed retrospectively. Preoperatively, phase I [non-invasive] and phase II [invasive] evaluation methods for epilepsy surgery were done. Seizure outcome was assessed with Engel's classification. The mean follow-up period was 27.2 months [from 12 months to 54 months]. Results : An MRI identified structural abnormality in eight patients and 3D-surface rendering revealed abnormal gyration in three. PET, SPECT, and surface EEG could not delineate the epileptogenic zone. Video-EEG monitoring with a subdural grid or depth electrodes verified the epileptogenic zone in all patients. Surgical procedures consisted of a resection of the SSMA and simultaneous callosotomy in two patients, a resection of the SSMA extending to the adjacent area in seven, a resection of a different area without a SSMA resection in seven, and a callosotomy in one. Seizure outcomes were class I in 11 [65%]. class II in five [29%], class III in one [6%]. Conclusion : In patients with intractable SSMA seizure, surgery was an excellent treatment modality. Precise delineation of the epileptogenic zone based on multimodal diagnostic methods can provide good surgical outcomes without neurological complications.