• Magazine of the Korean Society of Agricultural Engineers
• /
• v.26 no.2
• /
• pp.69-84
• /
• 1984

This study was carried out to investigate the time distribution of single storms and to establish the model of storm patterns in korea. Rainfall recording charts collected from 42 metheorological stations covering the Korean peninsula were analyzed. A single storm was defined as a rain period seperated from preceding and succeeding rainfall by 6 hours and more. Among the defined single storms, 1199 storms exceeding total rainfall of 80 mm were qualified for the study. Storm patterns were cklassified by quartile classification method and the relationship between cummulative percent of rainfalls and cummulative storm time was established for each quartile storm group. Time distribution models for each stations were prepared through the various analytical and inferential procedures. Obtained results are summarized as follows: 1. The percentile frequency of quartile storms for the first to the fourth quartile were 22.0%, 26.5%, 28.9% and 22.6%, respectively. The large variation of percentile frequency was show between the same quartile storms. The advanced type storm pattern was predominant in the west coastal type storm patterns predominantly when compared to the single storms with small total rainfalls. 3. The single storms with long storm durations tended to show delayed type storm patterns predominantly when compared to the single storms with short storm durations. 4. The percentile time distribution of quartile storms for 42 rin gaging stations was estimated. Large variations were observed between the percentiles of time distributions of different stations. 5. No significant differences were generally found between the time distribution of rainfalls with greater total rainfall and with less total rainfall. This fact suggests that the size of the total rainfall of single storms was not the main factor affecting the time distribution of heavy storms. 6. Also, no significant difference were found between the time distribution of rainfalls with long duration and with short duration. The fact indicates that the storm duration was no the main factor affecting the time distribution of heavy storms. 7. In Korea, among all single storms, 39.0% show 80 to 100mm of total rainfall which stands for the mode of the frequency distribution of total rainfalls. The median value of rainfalls for all single storms from the 42 stations was 108mm. The shape of the frequency distribution of total rainfalls showed right skewed features. No significant differences were shown in the shape of distribution histograms for total rainfall of quartile storms. The mode of rainfalls for the advanced type quartile storms was 80~100mm and their frequencies were 39~43% for respective quartiles. For the delayed type quartile storms, the mode was 80~100mm and their frequencies were 36!38%. 8. In Korea, 29% of all single storms show 720 to 1080 minutes of storm durations which was the highest frequency in the frequency distribution of storm durations. The median of the storm duration for all single storms form 42 stations was 1026 minutes. The shape of the frequency distribution was right skewed feature. For the advanced type storms, the higher frequency of occurrence was shown by the single storms with short durations, whereas for the delayed type quartile storms, the higher frequency was shown gy the long duration single storms. 9. The total rainfall of single storms was positively correlated to storm durations in all the stations throughout the nation. This fact was also true for most of the quartile storms. 10. The third order polynomial regression models were established for estimating the time distribution of quartile storms at different stations. The model test by relative error method resulted good agreements between estimated and observed values with the relative error of less than 0.10 in average.

• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.1-19
• /
• 2018

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people's movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven't been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image can be classified during fashion editorial task to classify and label the brand or style, and website query image can be processed by e-commerce multi-complex service providing item information or recommending similar item.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Asia-Pacific Journal of Business Venturing and Entrepreneurship
• /
• v.9 no.1
• /
• pp.15-23
• /
• 2014

A Case Study on Venture and Small-Business Executives' Use of Strategic Intuition in the Decision Making Process This paper is a case study on how Venture and Small-Business Executives managers can take advantage of their intuitions in situations where the business environment is increasingly uncertain, a novel situation occurs without any data to reflect on, when rational decision-making is not possible, and when the business environment changes. The case study is based on a literature review, in-depth interviews with 16 business managers, and an analysis of Klein, G's (1998) "Generic Mental Simulation Model." The "intuition" discussed in this analysis is classified into two types of intuition: the Expert Intuition which is based on one's own experiences, and Strategic Intuition which is based on the experience of others. Case study strategic management intuition and intuition, the experts were utilized differently. Features of professional intuition to work quickly without any effort by, while the strategic intuition, is time-consuming. Another feature that has already occurred, one expert intuition in decision-making about the widely used strategic intuition was used a lot in future decision-making. The case study results revealed that managers were using expert intuition and strategic intuition differentially. More specifically, Expert Intuition was activated effortlessly, while strategic intuition required more time. Also, expert intuition was used mainly for making judgments about events that have already happened, while strategic intuition was used more often for judgments regarding events in the future. The process of strategic intuition involved (1) Strategic concerns, (2) the discovery of medium, (3) Primary mental simulation, (4) The offsetting of key parameters, (5) secondary mental simulation, and (6) the decision making process. These steps were used to develop the "Strategic Intuition Decision-making Model" for Venture and Small-Business Executives. The case study results further showed that firstly, the success of decision-making was determined in the "secondary mental simulation' stage, and secondly, that more difficulty in management was encountered when expert intuition was used more than strategic intuition and lastly strategic intuition is possible to be educated.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.19 no.1
• /
• pp.95-107
• /
• 2017

In this study, a preliminary study was undertaken for development of a tunnel incident automatic detection system based on a machine learning algorithm which is to detect a number of incidents taking place in tunnel in real time and also to be able to identify the type of incident. Two road sites where CCTVs are operating have been selected and a part of CCTV images are treated to produce sets of training data. The data sets are composed of position and time information of moving objects on CCTV screen which are extracted by initially detecting and tracking of incoming objects into CCTV screen by using a conventional image processing technique available in this study. And the data sets are matched with 6 categories of events such as lane change, stoping, etc which are also involved in the training data sets. The training data are learnt by a resilience neural network where two hidden layers are applied and 9 architectural models are set up for parametric studies, from which the architectural model, 300(first hidden layer)-150(second hidden layer) is found to be optimum in highest accuracy with respect to training data as well as testing data not used for training. From this study, it was shown that the highly variable and complex traffic and incident features could be well identified without any definition of feature regulation by using a concept of machine learning. In addition, detection capability and accuracy of the machine learning based system will be automatically enhanced as much as big data of CCTV images in tunnel becomes rich.

• Journal of Korean Society of Transportation
• /
• v.23 no.7 s.85
• /
• pp.159-163
• /
• 2005

The propriety between suppliers and demanders in geometric design is very important. Although the final purpose of constructing roads is to concern about the driver s comfort, unfortunately, it has not been considered so far. We've considered the regularity and quickness in considering driver's comfort but there should be considered the safety for the accident as well. If drivers are appeared to be more speeding than designer's intention, there will be needed some supplements to increase the safety rate for the roads. Even if both an upward and downward section are supposed to exist at the same time for solid geometry of the roads like this, it is true that the recent design for the 3-D solid geometry section has been done as flat 2-D and the minimum plane curve radius and the maximum cant have been decided just by calculating without considering operating speed between an upward and downward section at the same point. In this investigation, thus, I'd like to calculate the safety of the cant by considering the speed features of the solid geometry for the first lane of four lane rural roads. To begin with, we investigated the driving speed of the car, which is not been influenced by a preceding car to analyze the influence of the geometrical structure by using Nc-97. Secondly, we statistically analyzed the driving features of the solid geometry after comparing the 6 sections, that is, measuring the driving speed feature at 12 points and combining the influence of the vertical geometry and plane geometry to the driving speed of the plane curve which was researched before. Finally, we estimated the value of cant which considers the driving speed not by using it which has applied uniformly without considering it properly, though there were some differences between a designed speed and driving speed through the result of the basic statistical analysis but by introducing the new safety rate rule, a notion of ${\alpha}$. As a result of the research, we could see the driving features of the car and suggest the safety rate which considers these. For considering the maximum cant, if we apply the safety rate, the result of this experiment, which considers 3-D solid geometry, there'll be the improvement of the driver's safety for designing roads. In addition, after collecting and analyzing the data for the road sections which have various geometrical structures by expanding this experiment it is considered that there should be developed the models which considers 3-D solid geometry.

• Asia pacific journal of information systems
• /
• v.18 no.2
• /
• pp.1-14
• /
• 2008

According to technology acceptance model(TAN) which is one of the most important research models for explaining IT users' behavior, on intention of using IT is determined by usefulness and ease of use of it. However, TAM wouldn't explain the performance of using IT while it has been considered as a very good model for prediction of the intention. Many people would not be confirmed in the performance of using IT until they can control it at their will, although they think it useful and easy to use. In other words, in addition to usefulness and ease of use as in TAM, controllability is also should be a factor to determine acceptance of IT. Especially, there is a very close relationship between controllability and ease of use, both of which explain the other sides of control over the performance of using IT, so called perceived behavioral control(PBC) in social psychology. The objective of this study is to identify the relationship between ease of use and controllability, and analyse the effects of both two beliefs over performance and intention in using IT. For this purpose, we review the issues related with PBC in information systems studies as well as social psychology, Based on a review of PBC, we suggest a research model which includes the relationship between control and performance in using IT, and prove its validity empirically. Since it was introduced as qa variable for explaining volitional control for actions in theory of planned behavior(TPB), there have been confusion about concept of PBC in spite of its important role in predicting so many kinds of actions. Some studies define PBC as self-efficacy that means actor's perception of difficulty or ease of actions, while others as controllability. However, this confusion dose not imply conceptual contradiction but a double-faced feature of PBC since the performance of actions is related with both self-efficacy and controllability. In other words, these two concepts are discriminated and correlated with each other. Therefore, PBC should be considered as a composite concept consisting of self-efficacy and controllability, Use of IT has been also one of important areas for predictions by PBC. Most of them have been studied by analysis of comparison in prediction power between TAM and TPB or modification of TAM by inclusion of PBC as another belief as like usefulness and ease of use. Interestingly, unlike the other applications in social psychology, it is hard to find such confusion in the concept of PBC in the studies for use of IT. In most of studies, controllability is adapted as PBC since the concept of self-efficacy is included in ease of use explicitly. Based on these discussions, we can suggest perceived use control(PUC) which is defined as perception of control over the performance of using IT and composed of controllability and ease of use as sub-concepts. We suggest a research model explaining acceptance of IT which includes the relationships of PUC with attitude and performance of using IT. For empirical test of our research model, two user groups are selected for surveying questionnaires. In the first group, there are freshmen who take a basic course for Microsoft Excel, and the second group consists of senior students who take a course for analysis of management information by Excel. Most of measurements are adapted ones that have been validated in the other studies, while performance is real score of mid-term in each class. In result, four hypotheses related with PUC are supported statistically with very low significance level. Main contribution of this study is suggestion of PUC through theoretical review of PBC. Specifically, a hierarchical model of PUC are derived from very rigorous studies in the relationship between self-efficacy and controllability with a view of PBC in social psychology. The relationship between PUC and performance is another main contribution.

• Journal of the Korean earth science society
• /
• v.28 no.2
• /
• pp.187-201
• /
• 2007

The first magnetotelluric (MT) transect across the Korean Peninsula was obtained traversing from the East Sea shoreline to the Yellow Sea shoreline. The MT survey profile was designed perpendicular to the strike of the principal geologic structure of the Korean Peninsula $(N30^{\circ}E)$, so-called 'China direction'. MT data were achieved at 50 sites with spacings of $3{\sim}8km$ along the 240 km survey line. The impedance responses are divided into four subsets reflecting typical geological units: the Kyonggi Massif, the Okchon Belt, the western part of the Kyongsang Basin, and the eastern part of the Kyongsang Basin. In the western part of the Kyongsang Basin, the thickness of the sedimentary layer is estimated to be about 3 km to 8 km and its resistivity is a few hundred ohm-m. A highly conductive layer with a resistivity of 1 to 30 ohm-m was detected beneath the sedimentary layer. The MT data at the Okchon Belt show peculiar responses with phase exceeding $90^{\circ}$. This feature may be explained by an electrically anisotropic structure which is composed of a narrow anisotropic block and an anisotropic layer. The Kyonggi Massif and the eastern part of Kyongsang Basin play a role of window to the deep geoelectrical structure because of the very high resistivity of upper crust. The second layers with highest resistivities in 1-D conductivity models occupy the upper crust with thicknesses of 13 km in the Kyonggi Massif and 18 km in the eastern Kyongsang Basin, respectively.

• Korean Journal of Psychosomatic Medicine
• /
• v.27 no.2
• /
• pp.101-110
• /
• 2019

Objectives : It is not clear which clinical variables are most closely associated with delirium in the Intensive Care Unit (ICU). By comparing clinical data of ICU delirium and non-delirium patients, we sought to identify variables that most effectively differentiate delirium from non-delirium. Methods : Medical records of 6,386 ICU patients were reviewed. Random Subset Feature Selection and Principal Component Analysis were utilized to select a set of clinical variables with the highest discriminatory capacity. Statistical analyses were employed to determine the separation capacity of two models-one using just the selected few clinical variables and the other using all clinical variables associated with delirium. Results : There was a significant difference between delirium and non-delirium individuals across 32 clinical variables. Richmond Agitation Sedation Scale (RASS), urinary catheterization, vascular catheterization, Hamilton Anxiety Rating Scale (HAM-A), Blood urea nitrogen, and Acute Physiology and Chronic Health Examination II most effectively differentiated delirium from non-delirium. Multivariable logistic regression analysis showed that, with the exception of vascular catheterization, these clinical variables were independent risk factors associated with delirium. Separation capacity of the logistic regression model using just 6 clinical variables was measured with Receiver Operating Characteristic curve, with Area Under the Curve (AUC) of 0.818. Same analyses were performed using all 32 clinical variables;the AUC was 0.881, denoting a very high separation capacity. Conclusions : The six aforementioned variables most effectively separate delirium from non-delirium. This highlights the importance of close monitoring of patients who received invasive medical procedures and were rated with very low RASS and HAM-A scores.

• Journal of Astronomy and Space Sciences
• /
• v.14 no.1
• /
• pp.94-108
• /
• 1997

The temperature of the upper thermosphere is generally varied with the solar activity, and largely with geomagnetic activity in the high latitude. The data analyzed in this study are acquired at two ground stations, Thule Air Base($76,5{deg} N, 68.4{deg} W, A = 86{deg}$) and $S{psi}ndre Str{psi}mfjord (67.0{deg} N, 50.9{deg} W, A = 74{deg}$), Greenland. Both stations are located in the high latitude not only geographically but also geomagnetically. The terrestrial night glow at 6300 ${angs}$ from atomic oxygen has been observed from the two ground-based Fabry-Perot interferometers, during periods of 1986~1991 in Thule Air Base and 1986~1994 in $S{psi}ndre Str{psi}mfjord$. Some features noted in this study are as follows: (1) The correlation between the solar activity and the measured thermospheric temperature is highest in the case of $3{leq}Kp{leq}4$ in Thule, and increases with the geomagnetic activity in $S{psi}ndre Str{psi}mfjord$. (2) The measured temperatures at Thule is generally higher than those at $S{psi}ndre Str{psi}mfjord$, but the latter shows steeper slope with the solar activity. (3) The harmonic analysis shows that the diurnal variation(24hrs) is the main feature of the daily temperature variation with a temperature peak at about 13-14 LT (LT=UT-4). However, the semi-diurnal variation is evident during the period of weak solar activity. (4) Generally the predicted temperatures from both MSIS86 and VSH models are lower than the measured temperature, and this discrepancy grows as the solar activity increases. Therefore, we urge modelers to develope a new thermospheric model utilizing broader sets of measurements, especially for high solar activity.