• Korean Journal of Remote Sensing
• /
• v.36 no.5_3
• /
• pp.1037-1051
• /
• 2020

The accurate monitoring and forecasting of the intensity of tropical cyclones (TCs) are able to effectively reduce the overall costs of disaster management. In this study, we proposed a multi-task learning (MTL) based deep learning model for real-time TC intensity estimation and forecasting with the lead time of 6-12 hours following the event, based on the fusion of geostationary satellite images and numerical forecast model output. A total of 142 TCs which developed in the Northwest Pacific from 2011 to 2016 were used in this study. The Communications system, the Ocean and Meteorological Satellite (COMS) Meteorological Imager (MI) data were used to extract the images of typhoons, and the Climate Forecast System version 2 (CFSv2) provided by the National Center of Environmental Prediction (NCEP) was employed to extract air and ocean forecasting data. This study suggested two schemes with different input variables to the MTL models. Scheme 1 used only satellite-based input data while scheme 2 used both satellite images and numerical forecast modeling. As a result of real-time TC intensity estimation, Both schemes exhibited similar performance. For TC intensity forecasting with the lead time of 6 and 12 hours, scheme 2 improved the performance by 13% and 16%, respectively, in terms of the root mean squared error (RMSE) when compared to scheme 1. Relative root mean squared errors(rRMSE) for most intensity levels were lessthan 30%. The lower mean absolute error (MAE) and RMSE were found for the lower intensity levels of TCs. In the test results of the typhoon HALONG in 2014, scheme 1 tended to overestimate the intensity by about 20 kts at the early development stage. Scheme 2 slightly reduced the error, resulting in an overestimation by about 5 kts. The MTL models reduced the computational cost about 300% when compared to the single-tasking model, which suggested the feasibility of the rapid production of TC intensity forecasts.

• Journal of the Korean Society of Radiology
• /
• v.14 no.2
• /
• pp.157-168
• /
• 2020

Advancements in segmentation methodology has made automatic segmentation of brain structures using structural images accurate and consistent. One method of automatic segmentation, which involves registering atlas information from template space to subject space, requires a high quality atlas with accurate boundaries for consistent segmentation. The Allen Mouse Brain Atlas, which has been widely accepted as a high quality reference of the mouse brain, has been used in various segmentations and can provide accurate coordinates and boundaries of mouse brain structures for tractography. Through probabilistic tractography, diffusion tensor images can be used to map comprehensive neuronal network of white matter pathways of the brain. Comparisons between neural networks of mouse and human brains showed that various clinical tests on mouse models were able to simulate disease pathology of human brains, increasing the importance of clinical mouse brain studies. However, differences between brain size of human and mouse brain has made it difficult to achieve the necessary image quality for analysis and the conditions for sufficient image quality such as a long scan time makes using live samples unrealistic. In order to secure a mouse brain image with a sufficient scan time, an Ex-vivo experiment of a mouse brain was conducted for this study. Using FSL, a tool for analyzing tensor images, we proposed a semi-automated segmentation and tractography analysis pipeline of the mouse brain and applied it to various mouse models. Also, in order to determine the useful signal-to-noise ratio of the diffusion tensor image acquired for the tractography analysis, images with various excitation numbers were compared.

• Journal of Korea Water Resources Association
• /
• v.47 no.2
• /
• pp.145-159
• /
• 2014

This study is to evaluate the future climate change impact on turbidity water and eutrophication for Chungju Lake by using CE-QUAL-W2 reservoir water quality model coupled with SWAT watershed model. The SWAT was calibrated and validated using 11 years (2000~2010) daily streamflow data at three locations and monthly stream water quality data at two locations. The CE-QUAL-W2 was calibrated and validated for 2 years (2008 and 2010) water temperature, suspended solid, total nitrogen, total phosphorus, and Chl-a. For the future assessment, the SWAT results were used as boundary conditions for CE-QUAL-W2 model run. To evaluate the future water quality variation in reservoir, the climate data predicted by MM5 RCM(Regional Climate Model) of Special Report on Emissions Scenarios (SRES) A1B for three periods (2013~2040, 2041~2070 and 2071~2100) were downscaled by Artificial Neural Networks method to consider Typhoon effect. The RCM temperature and precipitation outputs and historical records were used to generate pollutants loading from the watershed. By the future temperature increase, the lake water temperature showed $0.5^{\circ}C$ increase in shallow depth while $-0.9^{\circ}C$ in deep depth. The future annual maximum sediment concentration into the lake from the watershed showed 17% increase in wet years. The future lake residence time above 10 mg/L suspended solids (SS) showed increases of 6 and 17 days in wet and dry years respectively comparing with normal year. The SS occupying rate of the lake also showed increases of 24% and 26% in both wet and dry year respectively. In summary, the future lake turbidity showed longer lasting with high concentration comparing with present behavior. Under the future lake environment by the watershed and within lake, the future maximum Chl-a concentration showed increases of 19 % in wet year and 3% in dry year respectively.

• Journal of Intelligence and Information Systems
• /
• v.16 no.4
• /
• pp.213-225
• /
• 2010

Data mining has empowered the managers who are charge of the tasks in their company to present personalized and differentiated marketing programs to their customers with the rapid growth of information technology. Most studies on customer' response have focused on predicting whether they would respond or not for their marketing promotion as marketing managers have been eager to identify who would respond to their marketing promotion. So many studies utilizing data mining have tried to resolve the binary decision problems such as bankruptcy prediction, network intrusion detection, and fraud detection in credit card usages. The prediction of customer's response has been studied with similar methods mentioned above because the prediction of customer's response is a kind of dichotomous decision problem. In addition, a number of competitive data mining techniques such as neural networks, SVM(support vector machine), decision trees, logit, and genetic algorithms have been applied to the prediction of customer's response for marketing promotion. The marketing managers also have tried to classify their customers with quantitative measures such as recency, frequency, and monetary acquired from their transaction database. The measures mean that their customers came to purchase in recent or old days, how frequent in a period, and how much they spent once. Using segmented customers we proposed an approach that could enable to differentiate customers in the same rating among the segmented customers. Our approach employed support vector regression to forecast the purchase amount of customers for each customer rating. Our study used the sample that included 41,924 customers extracted from DMEF04 Data Set, who purchased at least once in the last two years. We classified customers from first rating to fifth rating based on the purchase amount after giving a marketing promotion. Here, we divided customers into first rating who has a large amount of purchase and fifth rating who are non-respondents for the promotion. Our proposed model forecasted the purchase amount of the customers in the same rating and the marketing managers could make a differentiated and personalized marketing program for each customer even though they were belong to the same rating. In addition, we proposed more efficient learning method by separating the learning samples. We employed two learning methods to compare the performance of proposed learning method with general learning method for SVRs. LMW (Learning Method using Whole data for purchasing customers) is a general learning method for forecasting the purchase amount of customers. And we proposed a method, LMS (Learning Method using Separated data for classification purchasing customers), that makes four different SVR models for each class of customers. To evaluate the performance of models, we calculated MAE (Mean Absolute Error) and MAPE (Mean Absolute Percent Error) for each model to predict the purchase amount of customers. In LMW, the overall performance was 0.670 MAPE and the best performance showed 0.327 MAPE. Generally, the performances of the proposed LMS model were analyzed as more superior compared to the performance of the LMW model. In LMS, we found that the best performance was 0.275 MAPE. The performance of LMS was higher than LMW in each class of customers. After comparing the performance of our proposed method LMS to LMW, our proposed model had more significant performance for forecasting the purchase amount of customers in each class. In addition, our approach will be useful for marketing managers when they need to customers for their promotion. Even if customers were belonging to same class, marketing managers could offer customers a differentiated and personalized marketing promotion.

• Proceedings of the Korea Society for Simulation Conference
• /
• 1996.05a
• /
• pp.1-3
• /
• 1996

시뮬레이션 기법을 이용한 시스템의 분석에 있어서 실험의 자동화는 현재 많은 연구와 개발이 진행 중인 분야이다. 컴퓨터와 정보통신 시스템에 대한 시뮬레이션의 예를 들어 보면, 수많은 모델을 대한 시뮬레이션을 수행할 경우 자동화된 실험의 제어가 요구되고 있다. 시뮬레이션 수행회수, 수행길이, 데이터 수집방법 등과 관련하여 시뮬레이션 실험방법이 자동화가 되지 않으면, 시뮬레이션 실험에 필요한 시간과 인적 자원이 상당히 커지게 되며 출력데이터에 대한 분석에 있어서도 어려움이 따르게 된다. 시뮬레이션 실험방법을 자동화하면서 효율적인 시뮬레이션 출력분석을 위해서는 시뮬레이션을 수행하는 경우에 항상 발생하는 초기편의 (initial bias)를 제거하는 문제가 선결되어야 한다. 시뮬레이션 출력분석에 사용되는 데이터들이 초기편의를 반영하지 않는 안정상태에서 수집된 것이어야만 실제 시스템에 대한 올바른 해석이 가능하다. 실제로 시뮬레이션 출력분석과 관련하여 가장 중요하면서도 어려운 문제는 시뮬레이션의 출력데이터가 이루는 추계적 과정 (stochastic process)의 안정상태 평균과 이 평균에 대한 신뢰구간(confidence interval: c. i.)을 구하는 것이다. 한 신뢰구간에 포함되어 있는 정보는 의사결정자에게 얼마나 정확하게 평균을 추정할 구 있는지 알려 준다. 그러나, 신뢰구간을 구성하는 일은 하나의 시뮬레이션으로부터 얻어진 출력데이터가 일반적으로 비정체상태(nonstationary)이고 자동상관(autocorrelated)되어 있기 때문에, 전통적인 통계적인 기법을 직접적으로 이용할 수 없다. 이러한 문제를 해결하기 위해 시뮬레이션 출력데이터 분석기법이 사용된다.본 논문에서는 초기편의를 제거하기 위해서 필요한 출력데이터의 제거시점을 찾는 새로운 기법으로, 유클리드 거리(Euclidean distance: ED)를 이용한 방법과 현재 패턴 분류(pattern classification) 문제에 널리 사용 중인 역전파 신경망(backpropagation neural networks: BNN) 알고리듬을 이용하는 방법을 제시한다. 이 기법들은 대다수의 기존의 기법과는 달리 시험수행(pilot run)이 필요 없으며, 시뮬레이션의 단일수행(single run) 중에 제거시점을 결정할 수 있다. 제거시점과 관련된 기존 연구는 다음과 같다. 콘웨이방법은 현재의 데이터가 이후 데이터의 최대값이나 최소값이 아니면 이 데이터를 제거시점으로 결정하는데, 알고기듬 구조상 온라인으로 제거시점 결정이 불가능하다. 콘웨이방법이 알고리듬의 성격상 온라인이 불가능한 반면, 수정콘웨이방법 (Modified Conway Rule: MCR)은 현재의 데이터가 이전 데이터와 비교했을 때 최대값이나 최소값이 아닌 경우 현재의 데이터를 제거시점으로 결정하기 때문에 온라인이 가능하다. 평균교차방법(Crossings-of-the-Mean Rule: CMR)은 누적평균을 이용하면서 이 평균을 중심으로 관측치가 위에서 아래로, 또는 아래서 위로 교차하는 회수로 결정한다. 이 기법을 사용하려면 교차회수를 결정해야 하는데, 일반적으로 결정된 교차회수가 시스템에 상관없이 일반적으로 적용가능하지 않다는 문제점이 있다. 누적평균방법(Cumulative-Mean Rule: CMR2)은 여러 번의 시험수행을 통해서 얻어진 출력데이터에 대한 총누적평균(grand cumulative mean)을 그래프로 그린 다음, 안정상태인 점을 육안으로 결정한다. 이 방법은 여러 번의 시뮬레이션을 수행에서 얻어진 데이터들의 평균들에 대한 누적평균을 사용하기 매문에 온라인 제거시점 결정이 불가능하며, 작업자가 그래프를 보고 임의로 결정해야 하는 단점이 있다. Welch방법(Welch's Method: WM)은 브라운 브리지(Brownian bridge) 통계량()을 사용하는데, n이 무한에 가까워질 때, 이 브라운 브리지 분포(Brownian bridge distribution)에 수렴하는 성질을 이용한다. 시뮬레이션 출력데이터를 가지고 배치를 구성한 후 하나의 배치를 표본으로 사용한다. 이 기법은 알고리듬이 복잡하고, 값을 추정해야 하는 단점이 있다. Law-Kelton방법(Law-Kelton's Method: LKM)은 회귀 (regression)이론에 기초하는데, 시뮬레이션이 종료된 후 누적평균데이터에 대해서 회귀직선을 적합(fitting)시킨다. 회귀직선의 기울기가 0이라는 귀무가설이 채택되면 그 시점을 제거시점으로 결정한다. 일단 시뮬레이션이 종료된 다음, 데이터가 모아진 순서의 반대 순서로 데이터를 이용하기 때문에 온라인이 불가능하다. Welch절차(Welch's Procedure: WP)는 5회이상의 시뮬레이션수행을 통해 수집한 데이터의 이동평균을 이용해서 시각적으로 제거시점을 결정해야 하며, 반복제거방법을 사용해야 하기 때문에 온라인 제거시점의 결정이 불가능하다. 또한, 한번에 이동할 데이터의 크기(window size)를 결정해야 한다. 지금까지 알아 본 것처럼, 기존의 방법들은 시뮬레이션의 단일 수행 중의 온라인 제거시점 결정의 관점에서는 미약한 면이 있다. 또한, 현재의 시뮬레이션 상용소프트웨어는 작업자로 하여금 제거시점을 임의로 결정하도록 하기 때문에, 실험중인 시스템에 대해서 정확하고도 정량적으로 제거시점을 결정할 수 없게 되어 있다. 사용자가 임의로 제거시점을 결정하게 되면, 초기편의 문제를 효과적으로 해결하기 어려울 뿐만 아니라, 필요 이상으로 너무 많은 양을 제거하거나 초기편의를 해결하지 못할 만큼 너무 적은 양을 제거할 가능성이 커지게 된다. 또한, 기존의 방법들의 대부분은 제거시점을 찾기 위해서 시험수행이 필요하다. 즉, 안정상태 시점만을 찾기 위한 시뮬레이션 수행이 필요하며, 이렇게 사용된 시뮬레이션은 출력분석에 사용되지 않기 때문에 시간적인 손실이 크게 된다.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_1
• /
• pp.1757-1766
• /
• 2021

The green coverage ratio is the ratio of the land area to green coverage area, and it is used as a practical urban greening index. The green coverage ratio is calculated based on the land cover map, but low spatial resolution and inconsistent production cycle of land cover map make it difficult to calculate the correct green coverage area and analyze the precise green coverage. Therefore, this study proposes a new method to calculate green coverage area using aerial images and deep neural networks. Green coverage ratio can be quickly calculated using manned aerial images acquired by local governments, but precise analysis is difficult because components of image such as acquisition date, resolution, and sensors cannot be selected and modified. This limitation can be supplemented by using an unmanned aerial vehicle that can mount various sensors and acquire high-resolution images due to low-altitude flight. In this study, we proposed a method to calculate green coverage ratio from manned or unmanned aerial images, and experimentally verified the proposed method. Aerial images enable precise analysis by high resolution and relatively constant cycles, and deep learning can automatically detect green coverage area in aerial images. Local governments acquire manned aerial images for various purposes every year and we can utilize them to calculate green coverage ratio quickly. However, acquired manned aerial images may be difficult to accurately analyze because details such as acquisition date, resolution, and sensors cannot be selected. These limitations can be supplemented by using unmanned aerial vehicles that can mount various sensors and acquire high-resolution images due to low-altitude flight. Accordingly, the green coverage ratio was calculated from the two aerial images, and as a result, it could be calculated with high accuracy from all green types. However, the green coverage ratio calculated from manned aerial images had limitations in complex environments. The unmanned aerial images used to compensate for this were able to calculate a high accuracy of green coverage ratio even in complex environments, and more precise green area detection was possible through additional band images. In the future, it is expected that the rust rate can be calculated effectively by using the newly acquired unmanned aerial imagery supplementary to the existing manned aerial imagery.

• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.175-197
• /
• 2021

Recently, with the development of deep learning technology, research on unstructured data analysis is being actively conducted, and it is showing remarkable results in various fields such as classification, summary, and generation. Among various text analysis fields, text classification is the most widely used technology in academia and industry. Text classification includes binary class classification with one label among two classes, multi-class classification with one label among several classes, and multi-label classification with multiple labels among several classes. In particular, multi-label classification requires a different training method from binary class classification and multi-class classification because of the characteristic of having multiple labels. In addition, since the number of labels to be predicted increases as the number of labels and classes increases, there is a limitation in that performance improvement is difficult due to an increase in prediction difficulty. To overcome these limitations, (i) compressing the initially given high-dimensional label space into a low-dimensional latent label space, (ii) after performing training to predict the compressed label, (iii) restoring the predicted label to the high-dimensional original label space, research on label embedding is being actively conducted. Typical label embedding techniques include Principal Label Space Transformation (PLST), Multi-Label Classification via Boolean Matrix Decomposition (MLC-BMaD), and Bayesian Multi-Label Compressed Sensing (BML-CS). However, since these techniques consider only the linear relationship between labels or compress the labels by random transformation, it is difficult to understand the non-linear relationship between labels, so there is a limitation in that it is not possible to create a latent label space sufficiently containing the information of the original label. Recently, there have been increasing attempts to improve performance by applying deep learning technology to label embedding. Label embedding using an autoencoder, a deep learning model that is effective for data compression and restoration, is representative. However, the traditional autoencoder-based label embedding has a limitation in that a large amount of information loss occurs when compressing a high-dimensional label space having a myriad of classes into a low-dimensional latent label space. This can be found in the gradient loss problem that occurs in the backpropagation process of learning. To solve this problem, skip connection was devised, and by adding the input of the layer to the output to prevent gradient loss during backpropagation, efficient learning is possible even when the layer is deep. Skip connection is mainly used for image feature extraction in convolutional neural networks, but studies using skip connection in autoencoder or label embedding process are still lacking. Therefore, in this study, we propose an autoencoder-based label embedding methodology in which skip connections are added to each of the encoder and decoder to form a low-dimensional latent label space that reflects the information of the high-dimensional label space well. In addition, the proposed methodology was applied to actual paper keywords to derive the high-dimensional keyword label space and the low-dimensional latent label space. Using this, we conducted an experiment to predict the compressed keyword vector existing in the latent label space from the paper abstract and to evaluate the multi-label classification by restoring the predicted keyword vector back to the original label space. As a result, the accuracy, precision, recall, and F1 score used as performance indicators showed far superior performance in multi-label classification based on the proposed methodology compared to traditional multi-label classification methods. This can be seen that the low-dimensional latent label space derived through the proposed methodology well reflected the information of the high-dimensional label space, which ultimately led to the improvement of the performance of the multi-label classification itself. In addition, the utility of the proposed methodology was identified by comparing the performance of the proposed methodology according to the domain characteristics and the number of dimensions of the latent label space.

• Korean Journal of Cognitive Science
• /
• v.33 no.1
• /
• pp.51-75
• /
• 2022

Transcranial direct current stimulation (tDCS) is a non-invasive brain stimulation that is able to alter neuronal activity in particular brain regions. Many studies have researched how tDCS modulates neuronal activity and reorganizes neural networks. However it is difficult to conclude the effect of brain stimulation because the studies are heterogeneous with respect to the stimulation parameter as well as individual difference. It is not fully in agreement with the effects of brain stimulation. In particular few studies have researched the reason of variability of brain stimulation in response to time so far. The study investigated individual variability of brain stimulation based on circadian rhythm and chronotype. Participants were divided into two groups which are morning type and evening type. The experiment was conducted by Zoom meeting which is video meeting programs. Participants were sent experiment tool which are Muse(EEG device), tdcs device, cell phone and cell phone holder after manuals for experimental equipment were explained. Participants were required to make a phone in frount of a camera so that experimenter can monitor online EEG data. Two participants who was difficult to use experimental devices experimented in a laboratory setting where experimenter set up devices. For all participants the accuracy of 98% was achieved by SVM using leave one out cross validation in classification in the the effects of morning stimulation and the evening stimulation. For morning type, the accuracy of 92% and 96% was achieved in classification in the morning stimulation and the evening stimulation. For evening type, it was 94% accuracy in classification for the effect of brain stimulation in the morning and the evening. Feature importance was different both in classification in the morning stimulation and the evening stimulation for morning type and evening type. Results indicated that the effect of brain stimulation can be explained with brain state and trait. Our study results noted that the tDCS protocol for target state is manipulated by individual differences as well as target state.

• Economic and Environmental Geology
• /
• v.55 no.6
• /
• pp.737-760
• /
• 2022

For the geological disposal of high-level radioactive wastes (HLW), an understanding of deep subsurface environment is essential through geological, hydrogeological, geochemical, and geotechnical investigations. Although South Korea plans the geological disposal of HLW, only a few studies have been conducted for characterizing the geochemistry of deep subsurface environment. To guide the hydrogeochemical research for selecting suitable repository sites, this study overviewed the status and trends in hydrogeochemical characterization of deep groundwater for the deep geological disposal of HLW in developed countries. As a result of examining the selection process of geological disposal sites in 8 countries including USA, Canada, Finland, Sweden, France, Japan, Germany, and Switzerland, the following geochemical parameters were needed for the geochemical characterization of deep subsurface environment: major and minor elements and isotopes (e.g., ³⁴S and ¹⁸O of SO₄^2-, ¹³C and ¹⁴C of DIC, ²H and ¹⁸O of water) of both groundwater and pore water (in aquitard), fracture-filling minerals, organic materials, colloids, and oxidation-reduction indicators (e.g., Eh, Fe²⁺/Fe³⁺, H₂S/SO₄^2-, NH₄⁺/NO₃^-). A suitable repository was selected based on the integrated interpretation of these geochemical data from deep subsurface. In South Korea, hydrochemical types and evolutionary patterns of deep groundwater were identified using artificial neural networks (e.g., Self-Organizing Map), and the impact of shallow groundwater mixing was evaluated based on multivariate statistics (e.g., M3 modeling). The relationship between fracture-filling minerals and groundwater chemistry also has been investigated through a reaction-path modeling. However, these previous studies in South Korea had been conducted without some important geochemical data including isotopes, oxidationreduction indicators and DOC, mainly due to the lack of available data. Therefore, a detailed geochemical investigation is required over the country to collect these hydrochemical data to select a geological disposal site based on scientific evidence.

• Journal of Korea Water Resources Association
• /
• v.56 no.6
• /
• pp.403-417
• /
• 2023

In this research, a methodology was developed for constructing an appropriate rainfall image database for estimating rainfall intensity based on CCTV video. The database was constructed in the Large-Scale Climate Environment Chamber of the Korea Conformity Laboratories, which can control variables with high irregularity and variability in real environments. 1,728 scenarios were designed under five different experimental conditions. 36 scenarios and a total of 97,200 frames were selected. Rain streaks were extracted using the k-nearest neighbor algorithm by calculating the difference between each image and the background. To prevent overfitting, data with pixel values greater than set threshold, compared to the average pixel value for each image, were selected. The area with maximum pixel variability was determined by shifting with every 10 pixels and set as a representative area (180×180) for the original image. After re-transforming to 120×120 size as an input data for convolutional neural networks model, image augmentation was progressed under unified shooting conditions. 92% of the data showed within the 10% absolute range of PBIAS. It is clear that the final results in this study have the potential to enhance the accuracy and efficacy of existing real-world CCTV systems with transfer learning.