• Korean Journal of Remote Sensing
• /
• v.34 no.2_1
• /
• pp.301-306
• /
• 2018

Urbanization causes urban floods and urban heat island in the summer, so it is necessary to understanding the changes of the thermal environment through urban climate and energy balance. This can be explained by the energy balance, but in urban areas, unlike the typical energy balance, the storage heat flux saved in the building or artificial land cover should be considered. Since the environment of each city is different, there is a difficulty in applying the method of retrieving the storage heat flux of the previous research. Especially, most of the previous studies are focused on the overseas cities, so it is necessary to study the storage heat retrieval suitable for various land cover and building characteristics of the urban areas in Korea. Therefore, the object of this study, it is to derive the regression formula which can quantitatively retrieve the storage heat using the data of the area where various surface types exist. To this end, nonlinear regression analysis was performed using net radiation and surface temperature data as independent variables and flux tower based storage heat estimates as dependent variables. The retrieved regression coefficients were applied to each independent variable to derive the storage heat retrieval regression formula. As a result of time series analysis with flux tower based storage heat estimates, it was well simulated high peak at day time and the value at night. Moreover storage heat retrieved in this study was possible continuous retrieval than flux tower based storage heat estimates. As a result of scatter plot analysis, accuracy of retrieved storage heat was found to be significant at $50.14Wm^{-2}$ and bias $-0.94Wm^{-2}$.

• The Journal of the Acoustical Society of Korea
• /
• v.35 no.4
• /
• pp.261-271
• /
• 2016

In the prediction of response of a pile in soil, numerical approaches such as a finite element method are generally applied due to complicate nonlinear behaviors of soils. However, the numerical methods based on the finite elements require heavy efforts in pile and soil modelling and also take long computing time. So their usage is limited especially in the early design stage in which principal dimensions and properties are not specified and tend to vary. On the contrary, theoretical approaches adopting linear approximations for soils are relatively simple and easy to model and take short computing time. Therefore, if they are validated to be reliable, they would be applicable in predicting responses of a pile in soil, particularly in early design stage. In case of wind turbines regarded in this study, it is required to assess their natural frequencies in early stages, and in this simulation the supporting pile inserted in soil could be replaced with a simplified elastic boundary condition at the bottom end of the wind turbine tower. To do this, analysis for a pile in soil is performed in this study to extract the spring constants at the top end of the pile. The pile in soil can be modelled as a beam on elastic spring by assuming that the soils deform within an elastic range. In this study, it is attempted to predict pile deformations and influence factors for lateral loads by means of the beam-on-spring model. As two example supporting structures for wind turbines, mono pile and suction pile models with different diameters are examined by evaluating their influence factors and validated by comparing them with those reported in literature. In addition, the deflection profiles along the depth and spring constants at the top end of the piles are compared to assess their supporting features.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.105-129
• /
• 2020

This study uses corporate data from 2012 to 2018 when K-IFRS was applied in earnest to predict default risks. The data used in the analysis totaled 10,545 rows, consisting of 160 columns including 38 in the statement of financial position, 26 in the statement of comprehensive income, 11 in the statement of cash flows, and 76 in the index of financial ratios. Unlike most previous prior studies used the default event as the basis for learning about default risk, this study calculated default risk using the market capitalization and stock price volatility of each company based on the Merton model. Through this, it was able to solve the problem of data imbalance due to the scarcity of default events, which had been pointed out as the limitation of the existing methodology, and the problem of reflecting the difference in default risk that exists within ordinary companies. Because learning was conducted only by using corporate information available to unlisted companies, default risks of unlisted companies without stock price information can be appropriately derived. Through this, it can provide stable default risk assessment services to unlisted companies that are difficult to determine proper default risk with traditional credit rating models such as small and medium-sized companies and startups. Although there has been an active study of predicting corporate default risks using machine learning recently, model bias issues exist because most studies are making predictions based on a single model. Stable and reliable valuation methodology is required for the calculation of default risk, given that the entity's default risk information is very widely utilized in the market and the sensitivity to the difference in default risk is high. Also, Strict standards are also required for methods of calculation. The credit rating method stipulated by the Financial Services Commission in the Financial Investment Regulations calls for the preparation of evaluation methods, including verification of the adequacy of evaluation methods, in consideration of past statistical data and experiences on credit ratings and changes in future market conditions. This study allowed the reduction of individual models' bias by utilizing stacking ensemble techniques that synthesize various machine learning models. This allows us to capture complex nonlinear relationships between default risk and various corporate information and maximize the advantages of machine learning-based default risk prediction models that take less time to calculate. To calculate forecasts by sub model to be used as input data for the Stacking Ensemble model, training data were divided into seven pieces, and sub-models were trained in a divided set to produce forecasts. To compare the predictive power of the Stacking Ensemble model, Random Forest, MLP, and CNN models were trained with full training data, then the predictive power of each model was verified on the test set. The analysis showed that the Stacking Ensemble model exceeded the predictive power of the Random Forest model, which had the best performance on a single model. Next, to check for statistically significant differences between the Stacking Ensemble model and the forecasts for each individual model, the Pair between the Stacking Ensemble model and each individual model was constructed. Because the results of the Shapiro-wilk normality test also showed that all Pair did not follow normality, Using the nonparametric method wilcoxon rank sum test, we checked whether the two model forecasts that make up the Pair showed statistically significant differences. The analysis showed that the forecasts of the Staging Ensemble model showed statistically significant differences from those of the MLP model and CNN model. In addition, this study can provide a methodology that allows existing credit rating agencies to apply machine learning-based bankruptcy risk prediction methodologies, given that traditional credit rating models can also be reflected as sub-models to calculate the final default probability. Also, the Stacking Ensemble techniques proposed in this study can help design to meet the requirements of the Financial Investment Business Regulations through the combination of various sub-models. We hope that this research will be used as a resource to increase practical use by overcoming and improving the limitations of existing machine learning-based models.

• The Korean Journal of Nuclear Medicine
• /
• v.32 no.3
• /
• pp.225-237
• /
• 1998

Purpose: We investigated the statistical methods to compose the functional brain map of human working memory and the principal factors that have an effect on the methods for localization. Materials and Methods: Repeated PET scans with successive four tasks, which consist of one control and three different activation tasks, were performed on six right-handed normal volunteers for 2 minutes after bolus injections of 925 MBq $H_2^{15}O$ at the intervals of 30 minutes. Image data were analyzed using SPM96 (Statistical Parametric Mapping) implemented with Matlab (Mathworks Inc., U.S.A.). Images from the same subject were spatially registered and were normalized using linear and nonlinear transformation methods. Significant difference between control and each activation state was estimated at every voxel based on the general linear model. Differences of global counts were removed using analysis of covariance (ANCOVA) with global activity as covariate. Using the mean and variance for each condition which was adjusted using ANCOVA, t-statistics was performed on every voxel To interpret the results more easily, t-values were transformed to the standard Gaussian distribution (Z-score). Results: All the subjects carried out the activation and control tests successfully. Average rate of correct answers was 95%. The numbers of activated blobs were 4 for verbal memory I, 9 for verbal memory II, 9 for visual memory, and 6 for conjunctive activation of these three tasks. The verbal working memory activates predominantly left-sided structures, and the visual memory activates the right hemisphere. Conclusion: We conclude that rCBF PET imaging and statistical parametric mapping method were useful in the localization of the brain regions for verbal and visual working memory.

• The Korean Journal of Nuclear Medicine
• /
• v.38 no.3
• /
• pp.241-252
• /
• 2004

Purpose: Population based structural and functional maps of the brain provide effective tools for the analysis and interpretation of complex and individually variable brain data. Brain MRI and PET standard templates and statistical probabilistic maps based on image data of Korean normal volunteers have been developed and probabilistic maps based on cytoarchitectonic data have been introduced. A quantification method using these data was developed for the objective assessment of regional intensity in the brain images. Materials and Methods: Age, gender and ethnic specific anatomical and functional brain templates based on MR and PET images of Korean normal volunteers were developed. Korean structural probabilistic maps for 89 brain regions and cytoarchitectonic probabilistic maps for 13 Brodmann areas were transformed onto the standard templates. Brain FDG PET and SPGR MR images of normal volunteers were spatially normalized onto the template of each modality and gender. Regional uptake of radiotracers in PET and gray matter concentration in MR images were then quantified by averaging (or summing) regional intensities weighted using the probabilistic maps of brain regions. Regionally specific effects of aging on glucose metabolism in cingulate cortex were also examined. Results: Quantification program could generate quantification results for single spatially normalized images per 20 seconds. Glucose metabolism change in cingulate gyrus was regionally specific: ratios of glucose metabolism in the rostral anterior cingulate vs. posterior cingulate and the caudal anterior cingulate vs. posterior cingulate were significantly decreased as the age increased. 'Rostral anterior'/'posterior' was decreased by 3.1% per decade of age ($P<10^{-11}$, r=0.81) and 'caudal anterior'/'posterior' was decreased by 1.7% ($P<10^{-8}$, r=0.72). Conclusion: Ethnic specific standard templates and probabilistic maps and quantification program developed in this study will be useful for the analysis of brain image of Korean people since the difference in shape of the hemispheres and the sulcal pattern of brain relative to age, gender, races, and diseases cannot be fully overcome by the nonlinear spatial normalization techniques.