• Journal of the Korean Institute of Landscape Architecture
• /
• v.37 no.4
• /
• pp.86-93
• /
• 2009

This study analyzes the systematic visual images and factors in and outside of the main courtyard in Byungsan-seowon. The results are as follows; In terms of space distribution, Ip-kyo-dang is located at an elevation of 85m and the distance to Byung-san is 365m. Byung-san with the mean gradient over $50^{\circ}$ looked so stiff, and the east side of that cliff is higher than west. In terms of the angle of elevation relationship between Man-dae-ru and Byung-san draw 10.5 degree and it suits with human scale. The D/H ratio of 1:3 makes the given place very spacious but the linear stiff shape of Byung-san may cause the feeling of closeness. The results of the visual image analysis of the main yard facing Byung-san is very positive with a score of 1.70 in openness, 1.78 in wideness, 1.96 in beauty, 1.96 in harmony for the spacious arrangement which overall, makes the seowon beautiful with many open spaces. There are 4 main implicated factors analyzed which are uniqueness, aesthetic, openness and nature. Out of the total variables, these factors' descriptive ability is 55.90% and the remaining 44.10% is error and peculiarities variables. The factor which contributed most to Byungsan-seowon's main yard's visual preference was the 'aesthetic' with B-values of 0.661 and 0.455 in the nature category.

• Journal of the Korean Society of Food Science and Nutrition
• /
• v.41 no.1
• /
• pp.136-142
• /
• 2012

Sogokju, a Korean glutinous rice wine and one of the oldest Korean traditional wines, is famous for its unique taste acquired from a 100-day incomplete fermentation process. This study investigated the effects of the degree of rice milling on the physicochemical and sensory characteristics of Sogokju. It evaluated the physicochemical characteristics, pasting and color properties, and structural properties of starch using four different degrees of milled rice (Oryza sativa L.) cultivar Dongjinchalbyeo. Samples of brown rice with milling yields of 92%, 84%, 76%, and 68% were produced using both abrasive and friction whiteners. This study showed that the protein, lipid, and ash content of milled rice decreased as the degree of milling increased. The lower hardness of the kernel below milling yield 92% suggested that milling may be related to the lower protein content of the kernel. The pasting curve showed a significant increase in viscosity properties as the degree of milling increased. This is due to the decrease in protein and lipid content, the increase in starch content, and the difference in amylopectin chain-length distribution. Further milling of white rice, based on 92% milling yield, had an effect on the amylopectin chain-length distribution due to the degree of polymerization (DPn) of 37~60. The long chain of amylopectin also contributed to the viscosity. The increase in the degree of milling decreased the glucose and total sugar content of Sogokju. However, it increased the total acidity of Sogokju. Moreover, the lightness of Sogokju decreased while its yellowness increased. These results indicate that the degree of milling can alter the taste and color of Sogokju. The sensory evaluation showed that the increase in the degree of milling decreased consumer preference for Sogokju. The sensory score for Sogokju was positively correlated with its brix degree, glucose content, pH, and protein content of raw rice.

• Journal of Distribution Science
• /
• v.13 no.10
• /
• pp.115-122
• /
• 2015

Purpose - This study was conducted to develop a beekeeping farm management standard checklist. This is essential to increase the competitive power of beekeeping farmers. Checklists in relation to crops and livestock were established by the Rural Development Administration in the 2000s. To date, 60 checklists have been created by crop and livestock experts. However, other farmers outside the 60 checklists are increasing. Therefore, extra development is required for these farmers. This study was conducted to meet farmers' requirements. The special farming dealt with in this study is beekeeping. Such checklists were not developed due to the small number of beekeeping farmers. However, these days, a number of such farmers are emerging. Research design, data, and methodology - Many related experts participated in this study. This study was conducted in four stages. First, a basic outline of beekeeping was created by surveying many kinds of beekeeping experts. The draft of the beekeeping checklist was created by a secondary advisory council. This draft was then sent to 14 beekeeping experts to confirm whether or not it was suitable as a management checklist. For collecting the experts' opinions, a direct visit survey was done through an arranged questionnaire. Additionally, a basic management checklist blueprint was reviewed by many experts. In the third stage, a Delphi survey method was utilized with a special Delphi questionnaire. In this stage, experts who participated in the first and second stages were excluded. As there were uncertain answers among them, a second Delphi survey was done. As a result of this survey, all answers were agreed among them. Results - From the results of this survey, four subjects in the management accomplishment index were determined. These are farming scale, average product per beehive, the sale price of honey (1kg), and the number of bee plates in the beehive. In the case of the management checklist content, five items were determined. These are beekeeping farming facilities, the environment around the farm land and general management, the product management of the beekeeping harvest, the management of the disease and pest, and farming management. This checklist will be utilized for beekeeping farmers to implement in a management situation. Conclusions - These days, the number of beekeeping farmers is increasing. The management checklist for beekeeping farmers will be used to improve their farming situation and marketing. Beekeeping farmers can understand their management by reviewing their checklist. After checking, the situation of management can be analyzed. Farmers can supplement weaknesses with expert advice. This checklist will be used by agricultural technique extension workers for farming management consulting. This checklist has to be complemented by a change in the management of the environment. This checklist will be delivered to beekeeping farmers after a verification survey is done. The result of the checklist score will be utilized for a benchmarking service to be implemented for beekeeping farmers to utilize.

• Korea journal of population studies
• /
• v.4 no.1
• /
• pp.105-130
• /
• 1981

This study was conducted to compare the characteristics of high performane areas for family planning with that of low performance areas and to find factors which strongly affected contraceptive practice behavior. For the study, eight areas were selected from 274 rural family planning canvassing areas of Korean Population Policy and Program Evaluation Study, which was an action study operated in all areas of Cheju Island from July 1, 1976 until December 31,1979. As a first step of the action study, Cheju Island was devided up 318 family planning canvasser areas Each area was consisted of 200 households in rural district and 300 households in urhan one Duriog the period of project, each canvassing area had been managed by a female family planning canvasser, selected by director of health center considering several individual conditions needed for family planning activities Basic activities of canvassers were to counsell all the eligihie couples in own charged area about family planning methods and also to distribute contraceptives such as condoms and oral pills. In case couples desire to accept sterilization including vasectomy and tubal-ligation, the canvassers played a linking role connecting potential client with family planning field workers. Canvassng areas shows significant differentce in performance for family planning, nevertheless they are supposed to have almost the same conditions regarding family planning distribution channel. Because the purpose of the Cheju project was to eliminate all the problems that existed in governmental distribution system, that is to remove geographic, economic, cognitive and administrative barriers Accumulated performances of family planning methods accepted by residents in each area were calculated by eligible women aged 14-49. And then canvassing areas were ranked according to performance score. Consequently, 4 areas in extremely high and low family planning performance areas were selected respectively. Major results were obtained by comparing characteristics of high performance area with that of low performance areas, which are as follows: 1. The mean number of living children was about the same both in high and low performance areas for family planning. But respondents' mean age (38.5) in high performance areas was higher than that (37.0) in low performance areas 2. Respondents' perception in the expectant educational level of others' children in high performance areas was higher than that in low performance areas, although respondents educational level, monthly expenditure and ratio of children in high school and above was not different. 3. Ratio of ownerships of TV and newspaper in high performance areas was highen than that in low performance areas 4. The duration of canvasser' charge in high performance areas was longer than that of low performance areas, showing the fact that canvassers didn't move cut in high performance areas 5. In high performance areas, canvassers' houses were relatively located in the center part of the village. And so villagers resided in near distances from the anvasser's house 6. 4H clubs' activities in high performance areas were more active than those in low performance areas Therefore it was assumed that cohesiveness of community in high performance areas were stronger than that in low areas. 7. Canvassers' family planning practice rate was higher than that in low performance areas, and also canvassers' human relationship was more sociable than that of canvassers in low performance areas. 8. Fourteen variables which showed relatively high significance level in $X^2$ and F test were selected as independent variables for stepwise regression analysis. According to the results of regression analysis. five of 14 variables-distributors education level ($R^2$=.4439), duration of distributor's charge ($R^2$=.6166), 4H club activities ($R^2$=.6697), canvasser's contraceptive practice ($R^2$=.7377) and location of distributions house ($R^2$=.8010) explained 80.1 percent of total variance.

• Journal of Gifted/Talented Education
• /
• v.20 no.3
• /
• pp.767-788
• /
• 2010

The purpose of this study was to examine emotional characteristics and to provide information about the special needs of counselling of science gifted in Korea. The subjects were 143 science gifted high school students in Busan that had been tested MBTI and MMPI-A. The distribution map of MBTI type was examined and Pearson's correlation, one-way ANOVA, multiple regression analysis were used to analyse the relation between MBTI and MMPI-A through SPSS 17.0 program. The results showed as follows: first, ENTP, INTP, ISTJ personality types and NT temperament type were the most frequently from the distribution map of MBTI type. Second, F1, F2, F, Hs, D, Pt, Sc and Si scales of MMPI-A were positively related to I preference of MBTI and K and Ma scales of MMPI-A were significantly related to E preference of MBTI from Pearson's correlation. Third, The score of IN group was significantly more high in F1, Hs, D, SC and Si scales of MMPI-A than other group in the relation between two combination preferences of MBTI and scale of MMPI-A. The following results were same; IS group in D, Si scales, EN group in Ma scale, IT group in Hs, D, Pt and S scales, IF group in VRIN, D and Si scales, ET in Ma scale, IJ group in D and Si, IP group in F1, F, Hs, D, Hy, Pt, Sc and Si scales, EJ and EP groups in Ma scale. Finally, I preference of MBTI by F1, F2, F, Hs, D, Pt, Sc and Si scales of MMPI-A, E preference of MBTI by Ma scale of MMPI-A, F preference of MBTI by K scale of MMPI-A and P preference of MBTI by Hy scale of MMPI-A were significantly predicted from multiple regression analysis. Limitations of the current study and the suggestions for further research were offered.

• Journal of Intelligence and Information Systems
• /
• v.25 no.3
• /
• pp.43-62
• /
• 2019

At one time, the anomaly detection sector dominated the method of determining whether there was an abnormality based on the statistics derived from specific data. This methodology was possible because the dimension of the data was simple in the past, so the classical statistical method could work effectively. However, as the characteristics of data have changed complexly in the era of big data, it has become more difficult to accurately analyze and predict the data that occurs throughout the industry in the conventional way. Therefore, SVM and Decision Tree based supervised learning algorithms were used. However, there is peculiarity that supervised learning based model can only accurately predict the test data, when the number of classes is equal to the number of normal classes and most of the data generated in the industry has unbalanced data class. Therefore, the predicted results are not always valid when supervised learning model is applied. In order to overcome these drawbacks, many studies now use the unsupervised learning-based model that is not influenced by class distribution, such as autoencoder or generative adversarial networks. In this paper, we propose a method to detect anomalies using generative adversarial networks. AnoGAN, introduced in the study of Thomas et al (2017), is a classification model that performs abnormal detection of medical images. It was composed of a Convolution Neural Net and was used in the field of detection. On the other hand, sequencing data abnormality detection using generative adversarial network is a lack of research papers compared to image data. Of course, in Li et al (2018), a study by Li et al (LSTM), a type of recurrent neural network, has proposed a model to classify the abnormities of numerical sequence data, but it has not been used for categorical sequence data, as well as feature matching method applied by salans et al.(2016). So it suggests that there are a number of studies to be tried on in the ideal classification of sequence data through a generative adversarial Network. In order to learn the sequence data, the structure of the generative adversarial networks is composed of LSTM, and the 2 stacked-LSTM of the generator is composed of 32-dim hidden unit layers and 64-dim hidden unit layers. The LSTM of the discriminator consists of 64-dim hidden unit layer were used. In the process of deriving abnormal scores from existing paper of Anomaly Detection for Sequence data, entropy values of probability of actual data are used in the process of deriving abnormal scores. but in this paper, as mentioned earlier, abnormal scores have been derived by using feature matching techniques. In addition, the process of optimizing latent variables was designed with LSTM to improve model performance. The modified form of generative adversarial model was more accurate in all experiments than the autoencoder in terms of precision and was approximately 7% higher in accuracy. In terms of Robustness, Generative adversarial networks also performed better than autoencoder. Because generative adversarial networks can learn data distribution from real categorical sequence data, Unaffected by a single normal data. But autoencoder is not. Result of Robustness test showed that he accuracy of the autocoder was 92%, the accuracy of the hostile neural network was 96%, and in terms of sensitivity, the autocoder was 40% and the hostile neural network was 51%. In this paper, experiments have also been conducted to show how much performance changes due to differences in the optimization structure of potential variables. As a result, the level of 1% was improved in terms of sensitivity. These results suggest that it presented a new perspective on optimizing latent variable that were relatively insignificant.

• Journal of Intelligence and Information Systems
• /
• v.27 no.1
• /
• pp.83-102
• /
• 2021

The government recently announced various policies for developing big-data and artificial intelligence fields to provide a great opportunity to the public with respect to disclosure of high-quality data within public institutions. KSURE(Korea Trade Insurance Corporation) is a major public institution for financial policy in Korea, and thus the company is strongly committed to backing export companies with various systems. Nevertheless, there are still fewer cases of realized business model based on big-data analyses. In this situation, this paper aims to develop a new business model which can be applied to an ex-ante prediction for the likelihood of the insurance accident of credit guarantee. We utilize internal data from KSURE which supports export companies in Korea and apply machine learning models. Then, we conduct performance comparison among the predictive models including Logistic Regression, Random Forest, XGBoost, LightGBM, and DNN(Deep Neural Network). For decades, many researchers have tried to find better models which can help to predict bankruptcy since the ex-ante prediction is crucial for corporate managers, investors, creditors, and other stakeholders. The development of the prediction for financial distress or bankruptcy was originated from Smith(1930), Fitzpatrick(1932), or Merwin(1942). One of the most famous models is the Altman's Z-score model(Altman, 1968) which was based on the multiple discriminant analysis. This model is widely used in both research and practice by this time. The author suggests the score model that utilizes five key financial ratios to predict the probability of bankruptcy in the next two years. Ohlson(1980) introduces logit model to complement some limitations of previous models. Furthermore, Elmer and Borowski(1988) develop and examine a rule-based, automated system which conducts the financial analysis of savings and loans. Since the 1980s, researchers in Korea have started to examine analyses on the prediction of financial distress or bankruptcy. Kim(1987) analyzes financial ratios and develops the prediction model. Also, Han et al.(1995, 1996, 1997, 2003, 2005, 2006) construct the prediction model using various techniques including artificial neural network. Yang(1996) introduces multiple discriminant analysis and logit model. Besides, Kim and Kim(2001) utilize artificial neural network techniques for ex-ante prediction of insolvent enterprises. After that, many scholars have been trying to predict financial distress or bankruptcy more precisely based on diverse models such as Random Forest or SVM. One major distinction of our research from the previous research is that we focus on examining the predicted probability of default for each sample case, not only on investigating the classification accuracy of each model for the entire sample. Most predictive models in this paper show that the level of the accuracy of classification is about 70% based on the entire sample. To be specific, LightGBM model shows the highest accuracy of 71.1% and Logit model indicates the lowest accuracy of 69%. However, we confirm that there are open to multiple interpretations. In the context of the business, we have to put more emphasis on efforts to minimize type 2 error which causes more harmful operating losses for the guaranty company. Thus, we also compare the classification accuracy by splitting predicted probability of the default into ten equal intervals. When we examine the classification accuracy for each interval, Logit model has the highest accuracy of 100% for 0~10% of the predicted probability of the default, however, Logit model has a relatively lower accuracy of 61.5% for 90~100% of the predicted probability of the default. On the other hand, Random Forest, XGBoost, LightGBM, and DNN indicate more desirable results since they indicate a higher level of accuracy for both 0~10% and 90~100% of the predicted probability of the default but have a lower level of accuracy around 50% of the predicted probability of the default. When it comes to the distribution of samples for each predicted probability of the default, both LightGBM and XGBoost models have a relatively large number of samples for both 0~10% and 90~100% of the predicted probability of the default. Although Random Forest model has an advantage with regard to the perspective of classification accuracy with small number of cases, LightGBM or XGBoost could become a more desirable model since they classify large number of cases into the two extreme intervals of the predicted probability of the default, even allowing for their relatively low classification accuracy. Considering the importance of type 2 error and total prediction accuracy, XGBoost and DNN show superior performance. Next, Random Forest and LightGBM show good results, but logistic regression shows the worst performance. However, each predictive model has a comparative advantage in terms of various evaluation standards. For instance, Random Forest model shows almost 100% accuracy for samples which are expected to have a high level of the probability of default. Collectively, we can construct more comprehensive ensemble models which contain multiple classification machine learning models and conduct majority voting for maximizing its overall performance.

• Journal of Global Scholars of Marketing Science
• /
• v.18 no.3
• /
• pp.1-25
• /
• 2008

Ⅰ. Introduction Retailers in the 21st century are being told that future retailers are those who can execute seamless multi-channel access. The reason is that retailers should be where shoppers want them, when they want them anytime, anywhere and in multiple formats. Multi-channel access is considered one of the top 10 trends of all business in the next decade (Patricia T. Warrington, et al., 2007) And most firms use both direct and indirect channels in their markets. Given this trend, we need to evaluate a channel equity more systematically than before as this issue is expected to get more attention to consumers as well as to brand managers. Consumers are becoming very much confused concerning the choice of place where they shop for durable goods as there are at least 6-7 retail options. On the other hand, manufacturers have to deal with category killers, their dealers network, Internet shopping malls, and other avenue of distribution channels and they hope their retail channel behave like extensions of their own companies. They would like their products to be foremost in the retailer's mind-the first to be proposed and effectively communicated to potential customers. To enable this hope to come reality, they should know each channel's advantages and disadvantages from consumer perspectives. In addition, customer satisfaction is the key determinant of retail customer loyalty. However, there are only a few researches regarding the effects of shopping satisfaction and perceptions on consumers' channel choices and channels. The purpose of this study was to assess Korean consumers' channel choice and satisfaction towards channels they prefer to use in the case of electronic goods shopping. Korean electronic goods retail market is one of good example of multi-channel shopping environments. As the Korea retail market has been undergoing significant structural changes since it had opened to global retailers in 1996, new formats such as hypermarkets, Internet shopping malls and category killers have arrived for the last decade. Korean electronic goods shoppers have seven major channels : (1)category killers (2) hypermarket (3) manufacturer dealer shop (4) Internet shopping malls (5) department store (6) TV home-shopping (7) speciality shopping arcade. Korean retail sector has been modernized with amazing speed for the last decade. Overall summary of major retail channels is as follows: Hypermarket has been number 1 retailer type in sales volume from 2003 ; non-store retailing has been number 2 from 2007 ; department store is now number 3 ; small scale category killers are growing rapidly in the area of electronics and office products in particular. We try to evaluate each channel's equity using a consumer survey. The survey was done by telephone interview with 1000 housewife who lives nationwide. Sampling was done according to 2005 national census and average interview time was 10 to 15 minutes. Ⅱ. Research Summary We have found that seven major retail channels compete with each other within Korean consumers' minds in terms of price and service. Each channel seem to have its unique selling points. Department stores were perceived as the best electronic goods shopping destinations due to after service. Internet shopping malls were perceived as the convenient channel owing to price checking. Category killers and hypermarkets were more attractive in both price merits and location conveniences. On the other hand, manufacturers dealer networks were pulling customers mainly by location and after service. Category killers and hypermarkets were most beloved retail channel for Korean consumers. However category killers compete mainly with department stores and shopping arcades while hypermarkets tend to compete with Internet and TV home shopping channels. Regarding channel satisfaction, the top 3 channels were service-driven retailers: department stores (4.27); dealer shop (4.21); and Internet shopping malls (4.21). Speciality shopping arcade(3.98) were the least satisfied channels among Korean consumers. Ⅲ. Implications We try to identify the whole picture of multi-channel retail shopping environments and its implications in the context of Korean electronic goods. From manufacturers' perspectives, multi-channel may cause channel conflicts. Furthermore, inter-channel competition draws much more attention as hypermarkets and category killers have grown rapidly in recent years. At the same time, from consumers' perspectives, 'buy where' is becoming an important buying decision as it would decide the level of shopping satisfaction. We need to develop the concept of 'channel equity' to manage multi-channel distribution effectively. Firms should measure and monitor their prime channel equity in regular basis to maximize their channel potentials. Prototype channel equity positioning map has been developed as follows. We expect more studies to develop the concept of 'channel equity' in the future.

• Journal of Distribution Science
• /
• v.9 no.4
• /
• pp.5-13
• /
• 2011

This study provides analysis of the effectiveness of domestic marketing strategies of the Korean coffee shop "Caffe Bene". It bases its evaluation on statistical outputs of 'choice attributes,' "market segmentation," demographic characteristics," and "satisfaction differences." The results are summarized in four points. First, five choice attributes were extracted from factor analysis: price, atmosphere, comfort, taste, and location; these are related to coffee shop selection behavior. Based on these five factors, cluster analysis was conducted, with statistical results classifying customers into three major groups: atmosphere oriented; comfort oriented; and taste oriented. Second, discriminant analysis tested cluster analysis and showed two discriminant functions: location and atmosphere. Third, cross-tabulation analysis based on demographic characteristics showed distinctive demographic characteristics within the three groups. Atmosphere oriented group, early-20s, as women of all ages was found to be 'walking down the street 'and 'through acquaintances' in many cases, as the cognitive path, and mostly found the store through 'outdoor advertising', and 'introduction'. Comfort oriented group was mainly women who are students in their early twenties or professionals, and appeared as a group to be very loyal because of high recommendation to other customers compared to other groups. Taste oriented group, unlike the other group, was mainly late-20s' college graduates, and was confirmed, as low loyalty, with lower recommendation activity. Fourth, to analyze satisfaction differences, one-way ANOVA was conducted. It shows that groups which show high satisfaction in the five main factors also show high menu satisfaction and high overall satisfaction. This results show that segmented marketing strategies are necessary because customers are considering price, atmosphere, comfort, taste, location when they choose coffee shop and demographics show different attributes based on segmented groups. For example, atmosphere oriented group is satisfied with shop interior and comfort while dissatisfied with price because most of the customers in this group are early 20s and do not have great financial capability. Thus, price discounting marketing strategies based on individual situations through CRM system is critical. Comfort oriented group shows high satisfaction level about location and shop comfort. Also, in this group, there are many early 20s female customers, students, and self-employed people. This group customers show high word of mouth tendency, hence providing positive brand image to the customers would be important. In case of taste oriented group, while the scores of taste and location are high, word of mouth score is low. This group is mainly composed of educated and professional many late 20s customers, therefore, menu differentiation, increasing quality of coffee taste and price discrimination is critical to increase customers' satisfaction. However, it is hard to generalize the results of study to other coffee shop brand, because this study have researched only one domestic coffee shop, Caffe Bene. Thus if future study expand the scope of locations, brands, and occupations, the results of the study would provide more generalizable results. Finally, research of customer satisfactions of menu, trust, loyalty, and switching cost would be critical in the future study.

• The Korean Journal of Nuclear Medicine
• /
• v.38 no.6
• /
• pp.486-491
• /
• 2004

Purpose: factor analysis and independent component analysis (ICA) has been used for handling dynamic image sequences. Theoretical advantages of a newly suggested ICA method, ensemble ICA, leaded us to consider applying this method to the analysis of dynamic myocardial $H_2^{15}O$ PET data. In this study, we quantified patients' blood flow using the ensemble ICA method. Materials and Methods: Twenty subjects underwent $H_2^{15}O$ PET scans using ECAT EXACT 47 scanner and myocardial perfusion SPECT using Vertex scanner. After transmission scanning, dynamic emission scans were initiated simultaneously with the injection of $555{\sim}740$ MBq $H_2^{15}O$. Hidden independent components can be extracted from the observed mixed data (PET image) by means of ICA algorithms. Ensemble learning is a variational Bayesian method that provides an analytical approximation to the parameter posterior using a tractable distribution. Variational approximation forms a lower bound on the ensemble likelihood and the maximization of the lower bound is achieved through minimizing the Kullback-Leibler divergence between the true posterior and the variational posterior. In this study, posterior pdf was approximated by a rectified Gaussian distribution to incorporate non-negativity constraint, which is suitable to dynamic images in nuclear medicine. Blood flow was measured in 9 regions - apex, four areas in mid wall, and four areas in base wall. Myocardial perfusion SPECT score and angiography results were compared with the regional blood flow. Results: Major cardiac components were separated successfully by the ensemble ICA method and blood flow could be estimated in 15 among 20 patients. Mean myocardial blood flow was $1.2{\pm}0.40$ ml/min/g in rest, $1.85{\pm}1.12$ ml/min/g in stress state. Blood flow values obtained by an operator in two different occasion were highly correlated (r=0.99). In myocardium component image, the image contrast between left ventricle and myocardium was 1:2.7 in average. Perfusion reserve was significantly different between the regions with and without stenosis detected by the coronary angiography (P<0.01). In 66 segment with stenosis confirmed by angiography, the segments with reversible perfusion decrease in perfusion SPECT showed lower perfusion reserve values in $H_2^{15}O$ PET. Conclusions: Myocardial blood flow could be estimated using an ICA method with ensemble learning. We suggest that the ensemble ICA incorporating non-negative constraint is a feasible method to handle dynamic image sequence obtained by the nuclear medicine techniques.