• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.139-156
• /
• 2021

The biggest reason for using a deep learning model in image classification is that it is possible to consider the relationship between each region by extracting each region's features from the overall information of the image. However, the CNN model may not be suitable for emotional image data without the image's regional features. To solve the difficulty of classifying emotion images, many researchers each year propose a CNN-based architecture suitable for emotion images. Studies on the relationship between color and human emotion were also conducted, and results were derived that different emotions are induced according to color. In studies using deep learning, there have been studies that apply color information to image subtraction classification. The case where the image's color information is additionally used than the case where the classification model is trained with only the image improves the accuracy of classifying image emotions. This study proposes two ways to increase the accuracy by incorporating the result value after the model classifies an image's emotion. Both methods improve accuracy by modifying the result value based on statistics using the color of the picture. When performing the test by finding the two-color combinations most distributed for all training data, the two-color combinations most distributed for each test data image were found. The result values were corrected according to the color combination distribution. This method weights the result value obtained after the model classifies an image's emotion by creating an expression based on the log function and the exponential function. Emotion6, classified into six emotions, and Artphoto classified into eight categories were used for the image data. Densenet169, Mnasnet, Resnet101, Resnet152, and Vgg19 architectures were used for the CNN model, and the performance evaluation was compared before and after applying the two-stage learning to the CNN model. Inspired by color psychology, which deals with the relationship between colors and emotions, when creating a model that classifies an image's sentiment, we studied how to improve accuracy by modifying the result values based on color. Sixteen colors were used: red, orange, yellow, green, blue, indigo, purple, turquoise, pink, magenta, brown, gray, silver, gold, white, and black. It has meaning. Using Scikit-learn's Clustering, the seven colors that are primarily distributed in the image are checked. Then, the RGB coordinate values of the colors from the image are compared with the RGB coordinate values of the 16 colors presented in the above data. That is, it was converted to the closest color. Suppose three or more color combinations are selected. In that case, too many color combinations occur, resulting in a problem in which the distribution is scattered, so a situation fewer influences the result value. Therefore, to solve this problem, two-color combinations were found and weighted to the model. Before training, the most distributed color combinations were found for all training data images. The distribution of color combinations for each class was stored in a Python dictionary format to be used during testing. During the test, the two-color combinations that are most distributed for each test data image are found. After that, we checked how the color combinations were distributed in the training data and corrected the result. We devised several equations to weight the result value from the model based on the extracted color as described above. The data set was randomly divided by 80:20, and the model was verified using 20% of the data as a test set. After splitting the remaining 80% of the data into five divisions to perform 5-fold cross-validation, the model was trained five times using different verification datasets. Finally, the performance was checked using the test dataset that was previously separated. Adam was used as the activation function, and the learning rate was set to 0.01. The training was performed as much as 20 epochs, and if the validation loss value did not decrease during five epochs of learning, the experiment was stopped. Early tapping was set to load the model with the best validation loss value. The classification accuracy was better when the extracted information using color properties was used together than the case using only the CNN architecture.

• Journal of Intelligence and Information Systems
• /
• v.26 no.4
• /
• pp.111-126
• /
• 2020

Most industries have recently become aware of the importance of customer lifetime value as they are exposed to a competitive environment. As a result, preventing customers from churn is becoming a more important business issue than securing new customers. This is because maintaining churn customers is far more economical than securing new customers, and in fact, the acquisition cost of new customers is known to be five to six times higher than the maintenance cost of churn customers. Also, Companies that effectively prevent customer churn and improve customer retention rates are known to have a positive effect on not only increasing the company's profitability but also improving its brand image by improving customer satisfaction. Predicting customer churn, which had been conducted as a sub-research area for CRM, has recently become more important as a big data-based performance marketing theme due to the development of business machine learning technology. Until now, research on customer churn prediction has been carried out actively in such sectors as the mobile telecommunication industry, the financial industry, the distribution industry, and the game industry, which are highly competitive and urgent to manage churn. In addition, These churn prediction studies were focused on improving the performance of the churn prediction model itself, such as simply comparing the performance of various models, exploring features that are effective in forecasting departures, or developing new ensemble techniques, and were limited in terms of practical utilization because most studies considered the entire customer group as a group and developed a predictive model. As such, the main purpose of the existing related research was to improve the performance of the predictive model itself, and there was a relatively lack of research to improve the overall customer churn prediction process. In fact, customers in the business have different behavior characteristics due to heterogeneous transaction patterns, and the resulting churn rate is different, so it is unreasonable to assume the entire customer as a single customer group. Therefore, it is desirable to segment customers according to customer classification criteria, such as loyalty, and to operate an appropriate churn prediction model individually, in order to carry out effective customer churn predictions in heterogeneous industries. Of course, in some studies, there are studies in which customers are subdivided using clustering techniques and applied a churn prediction model for individual customer groups. Although this process of predicting churn can produce better predictions than a single predict model for the entire customer population, there is still room for improvement in that clustering is a mechanical, exploratory grouping technique that calculates distances based on inputs and does not reflect the strategic intent of an entity such as loyalties. This study proposes a segment-based customer departure prediction process (CCP/2DL: Customer Churn Prediction based on Two-Dimensional Loyalty segmentation) based on two-dimensional customer loyalty, assuming that successful customer churn management can be better done through improvements in the overall process than through the performance of the model itself. CCP/2DL is a series of churn prediction processes that segment two-way, quantitative and qualitative loyalty-based customer, conduct secondary grouping of customer segments according to churn patterns, and then independently apply heterogeneous churn prediction models for each churn pattern group. Performance comparisons were performed with the most commonly applied the General churn prediction process and the Clustering-based churn prediction process to assess the relative excellence of the proposed churn prediction process. The General churn prediction process used in this study refers to the process of predicting a single group of customers simply intended to be predicted as a machine learning model, using the most commonly used churn predicting method. And the Clustering-based churn prediction process is a method of first using clustering techniques to segment customers and implement a churn prediction model for each individual group. In cooperation with a global NGO, the proposed CCP/2DL performance showed better performance than other methodologies for predicting churn. This churn prediction process is not only effective in predicting churn, but can also be a strategic basis for obtaining a variety of customer observations and carrying out other related performance marketing activities.

• Korean Journal of Ecology and Environment
• /
• v.34 no.1 s.93
• /
• pp.1-8
• /
• 2001

During April 1993 to November 1994, cations, anions, and conductivity were analyzed to examine how summer monsoon influences the ionic content of Taechung Reservoir, Korea. Interannual variability of ionic content reflected hydrological characteristics between the two years(high-flood year in 1993 vs. draught year in 1994). Cations, anions and conductivity were lowest during peak inflow in 1993 and highest during a drought in 1994. Floods in 1993 markedly decreased total salinity as a result of reduced Ca$^{2+}$ and HCO$_{3}\;^{-}$ and produced extreme spatial heterogeneity (i.e., longitudinal, vertical, and horizontal variation) in ionic concentrations. The dominant process modifying the longitudinal (the headwaters-to-downlake) and vertical (top-to-bottom) patterns in salinity was an interflow current during the 1993 monsoon. The interflow water plunged near a 27${\sim}$37 km-location (from the dam) of the mid-lake and passed through the 10${\sim}$30m stratum of the reservoir, resulting in an isolation of epilimnetic high conductivity water (>100 ${\mu}$S/cm) from advected river water with low conductivity (65${\sim}$75 ${\mu}$S/cm), During postmonsoon 1993, the factors regulating salinity differed spatially; salinity of downlake markedly declined as a result of dilution through the mixing of lake water with river water, whereas in the headwaters it increased due to enhanced CaCO$_{3}$ (originated from limestone/metamorphic rock) of groundwaters entering the reservoir. This result suggests an importance of the basin geology on ion compositions with hydrological characteristics. In 1994, salinity was markedly greater (p<0.001) relative to 1993 and ionic dilution did not occur during the monsoon due to reduced inflow. Overall data suggest that the primary factor influencing seasonal ionic concentrations and compositions in this system is the dilution process depending on the intensity of monsoon rainfall.

• Journal of the Korean Society for Marine Environment & Energy
• /
• v.16 no.1
• /
• pp.9-16
• /
• 2013

An experiment was conducted to evaluate the biochemical adverse effect of increased carbon dioxide in seawater on marine polychaete, Perinereis aibuhitensis. We measured the available energy reserves, Ea (total carbohydrate, protein, and lipid content) and the energy consumption, Ec (electron transport activity) of Perinereis aibuhitensis exposed for 7-d to a range of $CO_2$ concentration such as 0.39 (control =390 ppmv), 3.03 (=3,030 ppmv), 10.3 (=10,300 ppmv), and 30.1 (=30,100 ppmv) $CO_2$ mM, respectively. The cellular energy allocation (CEA) methodology was used to assess the adverse effects of toxic stress on the energy budget of the test organisms. The results of a decrease in CEA effect of increased carbon dioxide in seawater from all individual in Ea and Ec. Increase of carbon dioxide reduced pH in seawater, significantly. The chemical changes in sea- water caused by increasing $pCO_2$ might cause stresses to test organisms and changes in the cellular energy allocations. Results of this study can be used to understand the possible influence of $CO_2$ concentration increased by the leakage from sub-sea bed storage sites as well as fossil fuel combustion on marine organisms.

• Explosives and Blasting
• /
• v.30 no.2
• /
• pp.29-42
• /
• 2012

Variables influencing the free face movement due to rock blasting include the physical and mechanical properties, in particular the discontinuity characteristics, explosive type, charge weight, burden, blast-hole spacing, delay time between blast-holes or rows, stemming conditions. These variables also affects the blast vibration, air blast and size of fragmentation. For the design of surface blasting, the priority is given to the safety of nearby buildings. Therefore, blast vibration has to be controlled by analyzing the free face movement at the surface blasting sites and also blasting operation needs to be optimized to improve the fragmentation size. High-speed digital image analysis enables the analyses of the initial movement of free face of rock, stemming optimality, fragment trajectory, face movement direction and velocity as well as the optimal detonator initiation system. Even though The high-speed image analysis technique has been widely used in foreign countries, its applications can hardly be found in Korea. This thesis aims at carrying out a fundamental study for optimizing the blast design and evaluation using the high-speed digital image analysis. A series of experimentation were performed at two large surface blasting sites with the rock type of shale and granite, respectively. Emulsion and ANFO were the explosives used for the study. Based on the digital images analysis, displacement and velocity of the free face were scrutinized along with the analysis fragment size distribution. In addition, AUTODYN, 2-D FEM model, was applied to simulate detonation pressure, detonation velocity, response time for the initiation of the free face movement and face movement shape. The result show that regardless of the rock type, due to the displacement and the movement velocity have the maximum near the center of charged section the free face becomes curved like a bow. Compared with ANFO, the cases with Emulsion result in larger detonation pressure and velocity and faster reaction for the displacement initiation.

• Journal of the Korea Organic Resources Recycling Association
• /
• v.16 no.2
• /
• pp.47-56
• /
• 2008

The objective of this research is to evaluate the optimum recycling methods for the sewage sludge cakes at different microwave power-settings and for different periods of time. The dehydrated sewage sludge cakes used in this study was obtained from N wastewater treatment plan in the P City. The beginning drying processes were carried out in a microwave oven with 2,450 MHz frequency and power ranges of 1kW to 4 kW. The continuous conveyer drying system was also operated with 2,450 MHz frequency and power setting, ranging from of 1 kW to 6 kW. Initial moisture content of the sewage cake is 78~80%, and the moisture content decreased rapidly up to 0.2~2(wt%) within short periods due to breaking the cell walls. This study is also conducted to evaluate the characteristics of sewage sludge cakes with respect to important physical parameters effect on the thermal kinetics for evaporation water in the sludge which are operation times, moisture contents, drying rates, input amounts, flow rates and calorific values. It takes 60 minutes and 120 minutes to reach the critical moisture contents with power setting of 4 kW for 3kg/min and 6kg/min of the flow rates respectively. It takes 120 minutes and 110 minutes to reach the critical moisture contents with flow rates of 2.5 cm/min and sludge input of 6kg/min for the power settings of 4 kW and 6 kW respectively. The most effective value of the power for drying the sludge is 4 kW. Operation with 6kg/min and 4kW on 2cm of the sludge thickness can be effectively and inexpensively to reach the critical moisture contents, when you compare 2cm of the sludge thickness with 1cm and 3cm of the sludge thickness.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.35 no.6
• /
• pp.1309-1320
• /
• 2015

This study made an attempt to analyse the relationship between operation environment and fleet size per route which represents the level of service for inner-city bus service. Regression analysis method has been adopted as main analysis tool and 98 routes of bus operation status in Ulsan city as of 2013 has also been selected for analysis target. Correlation analysis was performed to identify the relationship between dependent and independent variables. There are three types of model for whole sample, type operation, and bus route operation system. These are the results of the current study. 1. The model developed for whole sample of 98 routes is as follows. Y(Fleet Size)=$-4.532+0.00002877*X_1$(Revenue). This model shows that it is necessary to have more than 140 passengers per day to increase fleet size of each bus route in Ulsan. 2. Models developed by type of operation (which are standard, express, and middle sized) are shown below. Stand Bus : Y(Fleet Size)=$-10.954+0.00004283*X_1$(Revenue). It is identified that more than 153 passengers need to use standard bus to increase fleet size per each standard bus, Middle Sized Bus : Y(Fleet Size)=-0.859+0.00001438*X1(Revenue). For middle sized bus, at least 52 daily passengers are needed to increase number of bus in each route. 3. Models developed for each route operation systems are as belows. Joint Operation Group : Y(Fleet Size)=$-4.786+0.00003028*X_1$(Revenue). Individual Operation Group : Y(Fleet Size)=$-2.339+0.00002030*X_1$(Revenue). These model provide similar result which 140 people is the minimum number of passenger to raise the number of vehicles in each route. This result shows that the route operation systems does not affect the raise number of cars significantly.

• Journal of Sensor Science and Technology
• /
• v.11 no.5
• /
• pp.309-318
• /
• 2002

A stoichiometric mixture of evaporating materials for $CdIn_2S_4$ single crystal thin films was prepared from horizontal electric furnace. To obtain the single crystal thin films, $CdIn_2S_4$ mixed crystal was deposited on thoroughly etched semi-insulating GaAs(100) substrate by the Hot Wall Epitaxy (HWE) system. The source and substrate temperatures were $630^{\circ}C$ and $420^{\circ}C$, respectively. The crystalline structure of the single crystal thin films was investigated by the photoluminescence and double crystal X-ray diffraction (DCXD). The carrier density and mobility of $CdIn_2S_4$ single crystal thin films measured with Hall effect by van der Pauw method are $9.01{\times}10^{16}\;cm^{-3}$ and $219\;cm^2/V{\cdot}s$ at 293 K, respectively. The temperature dependence of the energy band gap of the $CdIn_2S_4$ obtained from the absorption spectra was well described by the Varshni's relation, $E_g(T)=2.7116\;eV-(7.74{\times}10^{-4}\;eV)T^2/(T+434)$. The crystal field and the spin-orbit splitting energies for the valence band of the $CdIn_2S_4$ have been estimated to be 0.1291 eV and 0.0248 eV, respectively, by means of the photocurrent spectra and the Hopfield quasi cubic model. These results indicate that the splitting of the ${\Delta}so$ definitely exists in the ${\Gamma}5$ states of the valence band of the $AgInS_2$/GaAs epilayer. The three photocurrent peaks observed at 10K areascribed to the $A_1$-, $B_1$-, and C1-exciton peaks for n = 1.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.37 no.4
• /
• pp.32-41
• /
• 2009

Places such as famous urban sights and foreign settlements have recently been created in Korea, which reflects the increase in social and cultural exchanges and the number of place-making and landscape-planning projects. Understanding place identity is required in environmental design and planning. This paper examined the components of the identity of place through comparison of perceptions of "placeness" between residents and visitors of Seo-Rae Village, Seoul, Korea. More than 40% of French nationals in Korea reside in the village, which many Koreans visit as an urban sightseeing place. Twenty-five French inhabitants and forty-one Korean visitors were surveyed. Insideness of Seo-Rae Village was formed by limited activities and needs of the residents, primarily children's education and food purchases. Residents mentioned that the physical environment, especially buildings, lacked theauthenticity of French style. The symbolic meaning of the French village influenced the visitors' activities. The authenticity of French style, however, did not affect the perception of visitors and outsideness of the place. Visitors' activities and cognitive meanings of the place were formed mainly by the commercial environment established by Korean investments. Commercialization of the place and related activities played an important role in establishing insideness of the place. The village shows the phenomenon of the reversal of insideness and outsideness. The "placeness" of Seo-Rae village has been shaped by visitors' tastes rather than the true experience of insideness. Mass identity formed by mass media using French characteristics influenced this phenomenon. Insideness formed by inhabitants appeared to be a pseudo-place, which was created by the French tastes of visitors and lacked French authenticity. Seo-Rae Villagerepresents a superficial cloak.

• Progress in Medical Physics
• /
• v.17 no.3
• /
• pp.131-135
• /
• 2006

A commercial ion chamber matrix was examined the characteristics and its performance for radiotherapy qualify assurance. The device was the I'mRT 2D-MatriXX (Scanditronix-Wellhofer, Schwarzenbruck, Germany). The 2D-MatriXX device consists of a 1020 vented ion chamber array, arranged in $24{\times}24cm^2$ matrix. Each ion chamber has a volume of $0.08cm^3$, spacing of 0.762 cm and minimum sampling time of 20 ms. For the investigation of the characteristics, dose linearity, output factor, short-term reproducibility and dose rate dependency were tested. In the testing of dose linearity. It has shown a good signal linearity within 1% in the range of $1{\sim}800$cGy. Dose rate dependency was found to be lower than 0.4% (Range: 100-600 Mu/min) relative to a dose rate of 300 Mu/min as a reference. Output factors matched very well within 0.5% compared with commissioned beam data using a ionization chamber (CC01, Scanditronix-Wellhofer, Schwarzenbruck, Germany) in the range of field sizes $3{\times}3{\sim}24{\times}24cm^2$. Short-term reproducibility (6 times with a interval of 15 minute) was also shown a good agreement within 0.5%, when the temperature and the pressure were corrected by each time of measurement. in addition, we compared enhanced dynamic wedge (EDW, Varian, Palo Alto, USA) profiles from calculated values in the radiation planning system with those from measurements of the MatriXX. Furthermore, anon-uniform IMRT dose fluence was tested. All the comparison studies have shown good agreements. In this study, the MatriXX was evaluated as a reliable dosimeter, and it could be used as a simplistic and convenient tool for radiotherapy qualify assurance.