• Communications for Statistical Applications and Methods
• /
• 제30권5호
• /
• pp.453-465
• /
• 2023

The prediction problem of univariate records, though not addressed in multivariate records, has been discussed by many authors based on records values. There are various definitions for multivariate records among which depth-based records have been selected for the aim of this paper. In this paper, by means of the maximum likelihood and conditional median methods, point and interval predictions of depth values which are related to the future depth-based multivariate records are considered on the basis of the observed ones. The observations derived from some elements of the elliptical distributions are the main reason of studying this problem. Finally, the satisfactory performance of the prediction methods is illustrated via some simulation studies and a real dataset about Kermanshah city drought.

• KEPCO Journal on Electric Power and Energy
• /
• 제7권1호
• /
• pp.171-177
• /
• 2021

In terms of distribution planning, accurate electric load prediction is one of the most important factors. The future load prediction has manually been performed by calculating the maximum electric load considering loads transfer/switching and multiplying it with the load increase rate. In here, the risk of human error is inherent and thus an automated maximum electric load forecasting system is required. Although there are many existing methods and techniques to predict future electric loads, such as regression analysis, many of them have limitations in reflecting the nonlinear characteristics of the electric load and the complexity due to Photovoltaics (PVs), Electric Vehicles (EVs), and etc. This study, therefore, proposes a method of predicting future electric loads on distribution lines by using Machine Learning (ML) method that can reflect the characteristics of these nonlinearities. In addition, predictive models were developed based on actual data collected at KEPCO's existing distribution lines and the adequacy of developed models was verified as well. Also, as the distribution planning has a direct bearing on the investment, and amount of investment has a direct bearing on the maximum electric load, various baseline such as maximum, lowest, median value that can assesses the adequacy and accuracy of proposed ML based electric load prediction methods were suggested.

• 한국해양공학회지
• /
• 제33권6호
• /
• pp.526-534
• /
• 2019

The average coastline and the erosion control line introduced as the management coastline, and the average shoreline (MSL) was established from the observed coastline. Also, the median grain size and the wave height of 30-years return period were applied. The erosion control line (ECL) was established through the model, HaeSaBeeN. These two lines set the coastline for evaluation. Based on the observed monitoring data along the coastline, the 1-day variation according to the normal distribution was used to estimate the regional variation, and the width of the erosion was calculated by applying the median grain size (D₅₀) and the wave height of 30-years return period through the high-wave coastal erosion width model, i.e., HaeSaBeeN.

• Korean Journal of Hydrosciences
• /
• 제4권
• /
• pp.21-32
• /
• 1993

The estimates of both runoff depth and peak runoff by the basin runoff curve numbers, which are CN-II for antecedent moisture condition- II and CN -III for antecedent moisture condition-III, obtained from hydrological soil-cover complexes of 26 watersheds are investigated by making use of the observed curve numbers, which are median curve number and optimum curve number, computed from 250 rainfall-runoff records. For gaged basins the median curve numbers are recommended for the estimation of both runoff depth and peak runoff. For ungaged basin, found is that for the estimate of runoff depth CN-II is adequate and for peak runoff CN-II is suitable. Also investigated is the variation of the runoff curves during storms. By the variable runoff curve numbers, the prediction of runoff depth and peak runoff can be improved slightly.

• Tribology and Lubricants
• /
• 제10권3호
• /
• pp.62-70
• /
• 1994

An analytical model of the stress field caused by sliding indentation of brittle materials is developed. The complete stress field is treated as the superposition of applied normal and tangential forces with a sliding blister approximation of the localized inelastic deformation occuring just underneath the indenter. It is shown that lateral cracking is produced by the sliding blister stress field and that median cracking is caused by the applied contact forces. The model is combined with an experimental volume change measurements to show that the relative magnitude of tensile stresses governing lateral crack and median crack growth varies with the magnitude of the applied load. This prediction is consistent with the different regimes of experimentally observed cracking in soda-lime glass.

• 한국인공지능학회지
• /
• 제6권2호
• /
• pp.11-15
• /
• 2018

Breast cancer is one of the leading causes of cancer related death among women. So prediction of overall survival status is important into decided in adjuvant treatment. Deep belief network is a kind of artificial intelligence (AI). We intended to construct prediction model by deep belief network using associated clinicopathologic factors. 103881 cases were found in the Korean Breast Cancer Registry. After preprocessing of data, a total of 15733 cases were enrolled in this study. The median follow-up period was 82.4 months. In univariate analysis for overall survival (OS), the patients with advanced AJCC stage showed relatively high HR (HR=1.216 95% CI: 0.011-289.331, p=0.001). Based on results of univariate and multivariate analysis, input variables for learning model included 17 variables associated with overall survival rate. output was presented in one of two states: event or cencored. Individual sensitivity of training set and test set for predicting overall survival status were 89.6% and 91.2% respectively. And specificity of that were 49.4% and 48.9% respectively. So the accuracy of our study for predicting overall survival status was 82.78%. Prediction model based on Deep belief network appears to be effective in predicting overall survival status and, in particular, is expected to be applicable to decide on adjuvant treatment after surgical treatment.

• 천문학회보
• /
• 제46권1호
• /
• pp.50.2-51
• /
• 2021

Stars form exclusively in cold and dense molecular clouds. To fully understand star formation processes, it is hence a key to investigate how molecular clouds form out of the surrounding diffuse atomic gas. With an aim of shedding light in the process of the atomic-to-molecular transition in the interstellar medium, we analyze Arecibo HI emission and absorption spectral pairs along with TRAO/PMO 12CO(1-0) emission spectra toward 58 lines of sight probing in and around molecular clouds in the solar neighborhood, i.e., Perseus, Taurus, and California. 12CO(1-0) is detected from 19 out of 58 lines of sight, and we report the physical properties of HI (e.g., central velocity, spin temperature, and column density) in the vicinity of CO. Our preliminary results show that the velocity difference between the cold HI (Cold Neutral Medium or CNM) and CO (median ~ 0.7 km/s) is on average more than a factor of two smaller than the velocity difference between the warm HI (Warm Neutral Medium or WNM) and CO (median ~ 1.7 km/s). In addition, we find that the CNM tends to become colder (median spin temperature ~ 43 K) and abundant (median CNM fraction ~ 0.55) as it gets closer to CO. These results hints at the evolution of the CNM in the vicinity of CO, implying a close association between the CNM and molecular gas. Finally, in order to examine the role of HI in the formation of molecular gas, we compare the observed CNM properties to the theoretical model by Bialy & Sternberg (2016), where the HI column density for the HI-to-H2 transition point is predicted as a function of density, metallicity, and UV radiation field. Our comparison shows that while the model reproduces the observations reasonably well on average, the observed CNM components with high column densities are much denser than the model prediction. Several sources of this discrepancy, e.g., missing physical and chemical ingredients in the model such as the multi-phase ISM, non-equilibrium chemistry, and turbulence, will be discussed.

• Journal of Trauma and Injury
• /
• 제24권2호
• /
• pp.105-110
• /
• 2011

Purpose: Chest injuries in multiple trauma patients are major predisposing factor for increased length of stay in intensive care unit, prolonged mechanical ventilator, and respiratory complications such as pneumonia. The aim of this study is the evaluation of lung injury score as a risk factor for prolonged management in intensive care unit (ICU). Methods: Between June to August in 2011, 46 patients admitted to shock and trauma center in our hospital and 24 patients had associated chest damage without traumatic brain injury. Retrospectively, we calculated injury severity score (ISS), lung injury score, and the number of fractured ribs and performed nonparametric correlation analysis with length of stay in ICU and mechanical ventilator support. Results: Calculated lung injury score(<48 hours) was median 1(0-3) and ISS was median 30(8-38) in study population. They had median 2(0-14) fractured ribs. There were 2 bilateral fractures and 2 flail chest. Ventilator support was needed in 11(45.8%) of them for median 39 hours(6-166). The ISS of ventilator support group was median 34(24-34) and lung injury score was median 1.7(1.3-2.5). Tracheostomy was performed in one patient and it was only complicated case and ICU stay days was median 9(4-16). In correlation analysis, Lung injury score and ISS were significant with the length of stay in ICU but the number of fractured ribs and lung injury score were predicting factors for prolonged mechanical ventilator support. Conclusion: Lung injury score could be a possible prognostic factor for the prediction of increased length of stay in ICU and need for mechanical ventilator support.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제6권4호
• /
• pp.1248-1266
• /
• 2012

This paper presents a multilevel reversible data hiding method based on histogram shifting which can recover the original image losslessly after the hidden data has been extracted from the stego-image. The method of prediction is adopted in our proposed scheme and prediction errors are produced to explore the similarity of neighboring pixels. In this article, we propose two different predictors to generate the prediction errors, where the prediction is carried out using the center prediction method and the JPEG-LS median edge predictor (MED) to exploit the correlation among the neighboring pixels. Instead of the original image, these prediction errors are used to hide the secret information. Moreover, we also present an improved method to search for peak and zero pairs and also talk about the analogy of the same to improve the histogram shifting method for huge embedding capacity and high peak signal-to-noise ratio (PSNR). In the one-level hiding, our method keeps image qualities larger than 53 dB and the ratio of embedding capacity has 0.43 bpp (bit per pixel). Besides, the concept with multiple layer embedding procedure is applied for obtaining high capacity, and the performance is demonstrated in the experimental results. From our experimental results and analytical reasoning, it shows that the proposed scheme has higher PSNR and high data embedding capacity than that of other reversible data hiding methods presented in the literature.

• Bulletin of the Korean Chemical Society
• /
• 제33권2호
• /
• pp.613-619
• /
• 2012

We applied several machine learning methods for developing QSAR models for prediction of acute toxicity to fathead minnow. The multiple linear regression (MLR) and artificial neural network (ANN) method were applied to predict 96 h $LC_{50}$ (median lethal concentration) of 555 chemical compounds. Molecular descriptors based on 2D chemical structure were calculated by PreADMET program. The recursive partitioning (RP) model was used for grouping of mode of actions as reactive or narcosis, followed by MLR method of chemicals within the same mode of action. The MLR, ANN, and two RP-MLR models possessed correlation coefficients ($R^2$) as 0.553, 0.618, 0.632, and 0.605 on test set, respectively. The consensus model of ANN and two RP-MLR models was used as the best model on training set and showed good predictivity ($R^2$=0.663) on the test set.