• Korean small business review
• /
• v.31 no.2
• /
• pp.103-120
• /
• 2009

As business incubation centers (BICs) have been operating for more than 10 years in Korea, many early stage startups tend to use the services provided by the incubating centers. BICs in Korea have accumulated the knowledge and experience in the past ten years and their services have been considerably improved. The business incubating service has three facets : (1) business infrastructure service, (2) direct service, and (3) indirect service. The mission of BICs is to provide the early stage entrepreneurs with the incubating service in a limited period time to help them grow strong enough to survive the fierce competition after graduating from the incubation. However, the incubating services sometimes fail to foster the independence of new startup companies, and raise the dependence of many companies on BICs. Thus, the dependence on BICs is a very important factor to understand the survival of the incubated startup companies after graduation from BICs. The purpose of this study is to identify the main factors that influence the firm's dependence on BICs and to characterize the relationships among the identified factors. The business incubating service is a core construct of this study. It includes various activities and resources, such as offering the physical facilities, legal service, and connecting them with outside organizations. These services are extensive and take various forms. They are provided by BICs directly or indirectly. Past studies have identified various incubating services and classify them in different ways. Based on the past studies, we classify the business incubating service into three categories as mentioned above : (1) business infrastructure support, (2) direct support, and (3) networking support. The business infrastructure support is to provide the essential resources to start the business, such as physical facilities. The direct support is to offer the business resources available in the BICs, such as human, technical, and administrational resources. Finally, the indirect service was to support the resource in the outside of business incubation center. Dependence is generally defined as the degree to which a client firm needs the resources provided by the service provider in order to achieve its goals. Dependence is generated when a firm recognizes the benefits of interacting with its counterpart. Hence, the more positive outcomes a firm derives from its relationship with the partner, the more dependent on the partner the firm must inevitably become. In business incubating, as a resident firm is incubated in longer period, we can predict that her dependence on BICs would be stronger. In order to foster the independence of the incubated firms, BICs have to be able to manipulate the provision of their services to control the firms' dependence on BICs. Based on the above discussion, the research model for relationships between dependence and its affecting factors was developed. We surveyed the companies residing in BICs to test our research model. The instrument of our study was modified, in part, on the basis of previous relevant studies. For the purposes of testing reliability and validity, preliminary testing was conducted with firms that were residing in BICs and incubated by the BICs in the region of Gwangju and Jeonnam. The questionnaire was modified in accordance with the pre-test feedback. We mailed to all of the firms that had been incubated by the BICs with the help of business incubating managers of each BIC. The survey was conducted over a three week period. Gifts (of approximately ₩10,000 value) were offered to all actively participating respondents. The incubating period was reported by the business incubating managers, and it was transformed using natural logarithms. A total of 180 firms participated in the survey. However, we excluded 4 cases due to a lack of consistency using reversed items in the answers of the companies, and 176 cases were used for the analysis. We acknowledge that 176 samples may not be sufficient to conduct regression analyses with 5 research variables in our study. Each variable was measured through multiple items. We conducted an exploratory factor analysis to assess their unidimensionality. In an effort to test the construct validity of the instruments, a principal component factor analysis was conducted with Varimax rotation. The items correspond well to each singular factor, demonstrating a high degree of convergent validity. As the factor loadings for a variable (or factor) are higher than the factor loadings for the other variables, the instrument's discriminant validity is shown to be clear. Each factor was extracted as expected, which explained 70.97, 66.321, and 52.97 percent, respectively, of the total variance each with eigen values greater than 1.000. The internal consistency reliability of the variables was evaluated by computing Cronbach's alphas. The Cronbach's alpha values of the variables, which ranged from 0.717 to 0.950, were all securely over 0.700, which is satisfactory. The reliability and validity of the research variables are all, therefore, considered acceptable. The effects of dependence were assessed using a regression analysis. The Pearson correlations were calculated for the variables, measured by interval or ratio scales. Potential multicollinearity among the antecedents was evaluated prior to the multiple regression analysis, as some of the variables were significantly correlated with others (e.g., direct service and indirect service). Although several variables show the evidence of significant correlations, their tolerance values range between 0.334 and 0.613, thereby demonstrating that multicollinearity is not a likely threat to the parameter estimates. Checking some basic assumptions for the regression analyses, we decided to conduct multiple regression analyses and moderated regression analyses to test the given hypotheses. The results of the regression analyses indicate that the regression model is significant at p < 0.001 (F = 44.260), and that the predictors of the research model explain 42.6 percent of the total variance. Hypotheses 1, 2, and 3 address the relationships between the dependence of the incubated firms and the business incubating services. Business infrastructure service, direct service, and indirect service are all significantly related with dependence (β = 0.300, p < 0.001; β = 0.230, p < 0.001; β = 0.226, p < 0.001), thus supporting Hypotheses 1, 2, and 3. When the incubating period is the moderator and dependence is the dependent variable, the addition of the interaction terms with the antecedents to the regression equation yielded a significant increase in R2 (F change = 2.789, p < 0.05). In particular, direct service and indirect service exert different effects on dependence. Hence, the results support Hypotheses 5 and 6. This study provides several strategies and specific calls to action for BICs, based on our empirical findings. Business infrastructure service has more effect on the firm's dependence than the other two services. The introduction of an additional high charge rate for a graduated but allowed to stay in the BIC is a basic and legitimate condition for the BIC to control the firm's dependence. We detected the differential effects of direct and indirect services on the firm's dependence. The firms with long incubating period are more sensitive to indirect service positively, and more sensitive to direct service negatively, when assessing their levels of dependence. This implies that BICs must develop a strategy on the basis of a firm's incubating period. Last but not least, it would be valuable to discover other important variables that influence the firm's dependence in the future studies. Moreover, future studies to explain the independence of startup companies in BICs would also be valuable.

• Journal of the Korean Academy of Child and Adolescent Psychiatry
• /
• v.7 no.1
• /
• pp.77-91
• /
• 1996

In order to elucidate the biological etiology and the effects of comorbidity on biological variables in tic disorders, plasma serotonin (5-hydroxlfryptamine, 5-HT) and 5-hydroxy- indoleacetic acid (5-HIAA) we.e measured in 87 tic disorders and 30 control subjects. The 87 tic disorder were composed of 45 Tourette's disorder(TS), 22 chronic motor tic disorders (CMT) and 20 transient tic disorders (TTD). Among these patients,43 patients were pure tic disorder (PT), 28 subject also had attention deficit hyperactivity disorder (T+ADHD) and 16 subjects had obsessive compulsive disorders (T+ OCD) as comorbid disorders. The results are summarized as follows : 1) Plasma 5-HT levels showed significant positive correlations with plasma 5-HIAA levels (Pennon r=0.77, p<0.05). 2) Plasma 5-HT and 5-HIAA levels showed no significant correlation with age in tic disorders. 3) Plasma 5-HIAA and 5-HT levels showed no significant correlations with age in control subjects. 4) There was significant difference in plasma 5-HT levels among TS, CMT, TTD and control groups (ANOVA F=34.48, df=3, 113, p<0.01), and post-hoc test using Scheffe method showed significant differences between control and TS, control and CMT, control and ITD groups. But, post-hoc test showed no significant differences between TS and CMT, TS and TTD, CMT and TTD groups. 5) There was significant difference in plasma 5-HIAA levels among TS, CMT, TTD and control groups (ANOVA F=26.48, df=3, 113, p<0.01), and post-hoc test using Scheffe method showed significant differences between control and TS, control and CMT, control and TTD groups. But, post-hoc test showed no significant differences between TS and CMT, TS and TTD, CMT and TID groups.f) There was significant difference in plasma 5-HT and 5-HIAA levels among PT, T+ADHD, T+OCD and contol groups (ANOVA 5-HT, F=37.59, df=3, 113, p<0.01, 5-HIAA, F=27.37, df=3, 113, p<0.01), and post-hoc test using Scheffe method showed signiscant differences between control and PT, control and T+ADHD and control and T+OCB. But, post-hoc test showed no significant differences between PT and T+ADHD, PT and T+ OCD and T+ADHD and T+ OCD. These results show that decreased 5-HT and 5-HIAA levels may play a role in the genesis of tic disorders, but these findings have no significant correlations with the severity of tic disorders. And the comorbid disorders of tics may have minimal effects on the biochemical abnormalities. Future studies must be focused on the effects of serotonin agonists and antagonists on tic disorders and molecular biological methodology may enhance to elucidate the mechanisms of these abnormal findings.

• Journal of the Korean Academy of Child and Adolescent Psychiatry
• /
• v.11 no.2
• /
• pp.209-220
• /
• 2000

The purpose of this study is to find out the characteristics of depressive episode about major depression and bipolar disorder in child and adolescent. The subjects of this study were 34 major depression patients and 17 bipolar disorder patients hospitalized at child and adolescent psychiatry in OO university children's hospital from 1st March 1993 to 31st October 1999. The method of this study is to review socio-demographic characteristics, diagnostic classification, chief problems and symptoms at admission, frequency of symptoms, maternal pregnancy problem history, childhood developmental history, coexisting psychiatric disorders, family psychopathology and family history and therapeutic response through their chart. 1) The ratio of male was higher than that of female in major depressive disorder while they are similar in manic episode, bipolar disorder. 2) Average onset age of bipolar disorder was 14 years 1 month and it was 12 years 8 months in the case of major depression As a result, average onset age of major depression is lower than that of bipolar disorder. 3) The patients complained of vegetative symptoms than somatic symptoms in both bipolar disorder and depressive disorder. Also, the cases of major depression developed more suicide idea symptom while the case of bipolar disorder developed more aggressive symptoms. In the respect of psychotic symptoms, delusion was more frequently shown in major depression, but halucination was more often shown in bipolar disorder. 4) Anxiety disorder coexisted most frequently in two groups. And there coexisted symptoms such as somartoform disorder, mental retardation and personality disorder in both cases. 5) The influence of family loading was remarkable in both cases. Above all, the development of major depression had to do with child abuse history and inappropriate care of family. It is apparent that there are distinctive differences between major depression and bipolar disorder in child and adolescent through the study, just as in adult cases. Therefore the differences of clinical characteristics between two disorders is founded in coexisting disorders and clinical symptoms including onset age, somatic symptoms and vegetative symptoms.

• Radiation Oncology Journal
• /
• v.19 no.2
• /
• pp.127-135
• /
• 2001

Purpose : This study is a prospective randomized clinical trial comparing the efficacy and complication of anti-emetic drugs for prevention of nausea and vomiting after radiotherapy which has moderate emetogenic potential. The aim of this study was to investigate whether the anti-emetic efficacy of ondansetron $(Zofran^{\circledR})$ 8 mg bid dose (Group O) is better than the efficacy of metoclopramide 5 mg lid dose (Group M) in patients undergoing fractionated radiotherapy to the abdominal region. Materials and Methods : Study entry was restricted to those patients who met the following eligibility criteria: histologically confirmed malignant disease; no distant metastasis; performance status of not more than ECOG grade 2; no previous chemotherapy and radiotherapy. Between March 1997 and February 1998, 60 patients enrolled in this study. All patients signed a written statement of informed consent prior to enrollment. Blinding was maintained by dosing identical number of tablets including one dose of matching placebo for Group O. The extent of nausea, appetite loss, and the number of emetic episodes were recorded everyday using diary card. The mean score of nausea, appetite loss and the mean number of emetic episodes were obtained in a weekly interval. Results : Prescription error occurred in one patient. And diary cards have not returned in 3 patients due to premature refusal of treatment. Card from one patient was excluded from the analysis because she had a history of treatment for neurosis. As a result, the analysis consisted of 55 patients. Patient characteristics and radiotherapy characteristics were similar except mean age was $52.9{\pm}11.2$ in group M, $46.5{\pm}9.5$ in group O. The difference of age was statistically significant. The mean score of nausea, appetite loss and emetic episodes in a weekly interval was higher in group M than O. In group M, the symptoms were most significant at 5th week. In a panel data analysis using mixed procedure, treatment group was only significant factor detecting the difference of weekly score for all three symptoms. Ondansetron $(Zofran^{\circledR})$ 8 mg bid dose and metoclopramide 5 mg lid dose were well tolerated without significant side effects. There were no clinically important changes In vital signs or clinical laboratory parameters with either drug. Conclusion : Concerning the fact that patients with younger age have higher emetogenic potential, there are possibilities that age difference between two treatment groups lowered the statistical power of analysis. There were significant difference favoring ondansetron group with respect to the severity of nausea, vomiting and loss of appetite. We concluded that ondansetron is more effective anti-emetic agents in the control of radiotherapy-induced nausea, vomiting, loss of appetite without significant toxicity, compared with commonly used drug, i.e., metoclopramide. However, there were patients suffering emesis despite the administration of ondansetron. The possible strategies to improve the prevention and the treatment of radiotherapy-induced emesis must be further studied.

• The Korean Journal of Air & Space Law and Policy
• /
• v.32 no.1
• /
• pp.177-200
• /
• 2017

This writer have thus far reviewed the civil and criminal obligations of the directors of a parent company that sponsored financially troubled affiliates. What was discussed here applies to logistics companies in the same manner. Hanjin Shipping cannot expect its parent company, Korean Air to prop it up financially. If such financial aid is offered without any collateral, under Korean criminal law, the directors of the parent company bears the burden of civil and criminal responsibility. One way to get around this is to secure fairness in terms of the process and the content of aid. Fairness in terms of process refers to the board of directors making public all information and approving such aid. Fairness in terms of content refers to impartial transactions that block out any possibilities of the chairman of the corporate group acting in his private interest. In the case of Korean Air bailing out Hanjin, the meeting of board of directors were held five times and a thorough review was conducted on the risks involved in the loans being repaid or not. After the review, measures to guard against undesirable scenarios were established before finally deciding on bailing out Hanjin. As such, there are no issues. In terms of the fairness of content, too, there were practically no room for the majority shareholder or controlling shareholder to pocket profits at the expense of the company. This is because the continued aid offered to a financially troubled company (i.e. Hanjin Shipping) was a posing a burden to even the controlling shareholder. This writer argues that the concept of the interest of the entire corporate group needs to be recognized. That is, it must be recognized that the relationship of control and being controlled between parent company and affiliate company, or between affiliate companies serves a practical benefit to the ongoing concern and growth of the group and is therefore just. Moreover, the corporate group and its affiliates, as well as their directors and management must recognize that they have an obligation to prioritize the interests of the corporate group ahead of the interests of the company that they are directly associated with. As such, even if Korean Air offered a loan to Hanjin Shipping without collateral, the act cannot be treated as an offense to law, nor can the directors be accused of damages that they bear the responsibility of compensating under civil law.

• Korean Journal of Fisheries and Aquatic Sciences
• /
• v.25 no.2
• /
• pp.79-92
• /
• 1992

This study was carried out to develop new techniques useful for cryopreservation, thawing and artificial insemination, and ultrastructural changes of cryopreserved spermatozoa in rainbow trout(Oncorhynchus mykiss) . Two extenders, such as Tyrode solution and Whittingham's $T_6$ solution, were used to preserve rainbow trout sperm in refrigerator $(-20,\;-40\;and\;-70^{\circ}C)$ or liquid nitrogen $%(-196^{\circ})$. Hand-stripped semen was diluted to 1:16 with two extenders, an then the semen were frozen after mixing semen and each extender containing 1M or 1.5M DMSO solution to 1:1. After 60 days cryopreserved semen was thawed in a $13^{\circ}$ water bath, and subsequently centrifugated. After centrifugation at 1,000 rpm for 5 min thawed semen was washed with extenders, and then fertilized with fresh eggs. The results obtained in these experiments were summarized as follows: After cryopreservation, over 75% of spermatozoa were appeared motile and the survival rate was high. Following cryopreservation by the addition of cryoprotectant such as DMSO, methanol and glycerol, the fertilization rate of the thawed spermatozoa appeared over $99\%$ compared with the control having $99\%$ of fertilization rate. There was no difference between the control and experimental groups such as $(-20^{\circ}C\;-40^{\circ}C\;and\;-70^{\circ}C)$ and $-196^{\circ}$ in fertilization rate. Following cryopreservation at $-196^{\circ}$ by the addition of 1M DMSO of cryoprotectant, each fertilization rate following 24 hours and hatching rate following 24 days showed $96\%$ and $8\%$ by the addition of BSA, but showed $98\%\;and\;10%$ by no addition of BSA. Following 2 months of cryopreservation by the addition of 1M DMSO of cryoprotectant, there were $10%$ of hatching rate at $-196^{\circ}\;and\;10\%\;and\;35\%,\;respectively,\;at\;-40^{\circ}C\;and\;-70^{\circ}C$. Following 2 months of cryopreservation by the addition of 1M methanol of cryoprotectant, there were $22\%$ of fertilization rate at $-20^{\circ}C,\;and\;28\%,\;at\;-70^{\circ}C$ Following 2 months of cryopreservation by the addition of 1M glycerol of cryoprotectant, there were $22\%$ of fertilization rate at $-20^{\circ}C$, and $33\%,\;at\;-70^{\circ}C$. pollowing 2 months of cryopreservation by the addition of 1.5M DMSO of cryoprotectant, there were $27\%$ of fertilization rate at $-20^{\circ}C,\;an\;36\%\;and \;35\%,\;respectively,\;at\;-40^{\circ}C\;and\;-70^{\circ}C$. Following 2 months of cryopreservation by the addition of 1.5M glycerol of cryoprotectant, there were $34\% \;of\;fertilization\;rate\;at\;-20^{\circ}C, \;and\;31\%\;and\;31\%,\;respectively,\;at \;-40^{\circ}C\;and\;-70^{\circ}$. Following 2 months of cryopreservation by the addition of 1.5M methanol of cryoprotectant, there were $28\%$ of fertilization rate at $-20^{\circ}C,\;and\;29\%\;and\;28\%,\;respectively,\;at\;-40^{\circ}C\;and\;-70^{\circ}C.$ From 10 days and 15 days following fertilization at $13^{\circ}C\;and\;10^{\circ}C$, respectively, the mortality rate of fertilized ova was markedly increased. The middle piece of spermatozoa had two set of central doublets, nine set of outer coarse fibres, and mitochondrial sheath. Spermatozoa went through morphological changes during storage, e.g. winding of flagella, detachment of the nuclear envelope and the plasma membrane from the nucleus of the sperm head. There were $1\%$ abnormal spermatozoa in fresh sperm and about $15\%$ during storage.

• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.167-181
• /
• 2018

Over the past decade, deep learning has been in spotlight among various machine learning algorithms. In particular, CNN(Convolutional Neural Network), which is known as the effective solution for recognizing and classifying images or voices, has been popularly applied to classification and prediction problems. In this study, we investigate the way to apply CNN in business problem solving. Specifically, this study propose to apply CNN to stock market prediction, one of the most challenging tasks in the machine learning research. As mentioned, CNN has strength in interpreting images. Thus, the model proposed in this study adopts CNN as the binary classifier that predicts stock market direction (upward or downward) by using time series graphs as its inputs. That is, our proposal is to build a machine learning algorithm that mimics an experts called 'technical analysts' who examine the graph of past price movement, and predict future financial price movements. Our proposed model named 'CNN-FG(Convolutional Neural Network using Fluctuation Graph)' consists of five steps. In the first step, it divides the dataset into the intervals of 5 days. And then, it creates time series graphs for the divided dataset in step 2. The size of the image in which the graph is drawn is $40(pixels){\times}40(pixels)$, and the graph of each independent variable was drawn using different colors. In step 3, the model converts the images into the matrices. Each image is converted into the combination of three matrices in order to express the value of the color using R(red), G(green), and B(blue) scale. In the next step, it splits the dataset of the graph images into training and validation datasets. We used 80% of the total dataset as the training dataset, and the remaining 20% as the validation dataset. And then, CNN classifiers are trained using the images of training dataset in the final step. Regarding the parameters of CNN-FG, we adopted two convolution filters ($5{\times}5{\times}6$ and $5{\times}5{\times}9$) in the convolution layer. In the pooling layer, $2{\times}2$ max pooling filter was used. The numbers of the nodes in two hidden layers were set to, respectively, 900 and 32, and the number of the nodes in the output layer was set to 2(one is for the prediction of upward trend, and the other one is for downward trend). Activation functions for the convolution layer and the hidden layer were set to ReLU(Rectified Linear Unit), and one for the output layer set to Softmax function. To validate our model - CNN-FG, we applied it to the prediction of KOSPI200 for 2,026 days in eight years (from 2009 to 2016). To match the proportions of the two groups in the independent variable (i.e. tomorrow's stock market movement), we selected 1,950 samples by applying random sampling. Finally, we built the training dataset using 80% of the total dataset (1,560 samples), and the validation dataset using 20% (390 samples). The dependent variables of the experimental dataset included twelve technical indicators popularly been used in the previous studies. They include Stochastic %K, Stochastic %D, Momentum, ROC(rate of change), LW %R(Larry William's %R), A/D oscillator(accumulation/distribution oscillator), OSCP(price oscillator), CCI(commodity channel index), and so on. To confirm the superiority of CNN-FG, we compared its prediction accuracy with the ones of other classification models. Experimental results showed that CNN-FG outperforms LOGIT(logistic regression), ANN(artificial neural network), and SVM(support vector machine) with the statistical significance. These empirical results imply that converting time series business data into graphs and building CNN-based classification models using these graphs can be effective from the perspective of prediction accuracy. Thus, this paper sheds a light on how to apply deep learning techniques to the domain of business problem solving.

• 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.24 no.2
• /
• pp.221-241
• /
• 2018

Deep learning is getting attention recently. The deep learning technique which had been applied in competitions of the International Conference on Image Recognition Technology(ILSVR) and AlphaGo is Convolution Neural Network(CNN). CNN is characterized in that the input image is divided into small sections to recognize the partial features and combine them to recognize as a whole. Deep learning technologies are expected to bring a lot of changes in our lives, but until now, its applications have been limited to image recognition and natural language processing. The use of deep learning techniques for business problems is still an early research stage. If their performance is proved, they can be applied to traditional business problems such as future marketing response prediction, fraud transaction detection, bankruptcy prediction, and so on. So, it is a very meaningful experiment to diagnose the possibility of solving business problems using deep learning technologies based on the case of online shopping companies which have big data, are relatively easy to identify customer behavior and has high utilization values. Especially, in online shopping companies, the competition environment is rapidly changing and becoming more intense. Therefore, analysis of customer behavior for maximizing profit is becoming more and more important for online shopping companies. In this study, we propose 'CNN model of Heterogeneous Information Integration' using CNN as a way to improve the predictive power of customer behavior in online shopping enterprises. In order to propose a model that optimizes the performance, which is a model that learns from the convolution neural network of the multi-layer perceptron structure by combining structured and unstructured information, this model uses 'heterogeneous information integration', 'unstructured information vector conversion', 'multi-layer perceptron design', and evaluate the performance of each architecture, and confirm the proposed model based on the results. In addition, the target variables for predicting customer behavior are defined as six binary classification problems: re-purchaser, churn, frequent shopper, frequent refund shopper, high amount shopper, high discount shopper. In order to verify the usefulness of the proposed model, we conducted experiments using actual data of domestic specific online shopping company. This experiment uses actual transactions, customers, and VOC data of specific online shopping company in Korea. Data extraction criteria are defined for 47,947 customers who registered at least one VOC in January 2011 (1 month). The customer profiles of these customers, as well as a total of 19 months of trading data from September 2010 to March 2012, and VOCs posted for a month are used. The experiment of this study is divided into two stages. In the first step, we evaluate three architectures that affect the performance of the proposed model and select optimal parameters. We evaluate the performance with the proposed model. Experimental results show that the proposed model, which combines both structured and unstructured information, is superior compared to NBC(Naïve Bayes classification), SVM(Support vector machine), and ANN(Artificial neural network). Therefore, it is significant that the use of unstructured information contributes to predict customer behavior, and that CNN can be applied to solve business problems as well as image recognition and natural language processing problems. It can be confirmed through experiments that CNN is more effective in understanding and interpreting the meaning of context in text VOC data. And it is significant that the empirical research based on the actual data of the e-commerce company can extract very meaningful information from the VOC data written in the text format directly by the customer in the prediction of the customer behavior. Finally, through various experiments, it is possible to say that the proposed model provides useful information for the future research related to the parameter selection and its performance.

• Tuberculosis and Respiratory Diseases
• /
• v.41 no.5
• /
• pp.521-530
• /
• 1994

Background : The past studies on prediction formulas of pulmonary function parameters in healthy nonsmoking Korean adults have been performed in relatively small number of subjects and the reported results were restricted on a few parameters. Also there was no systematic investigation into the effect of smoking on pulmonary function parameters in smokers who have no respiratory symptoms. Therefore we attempted to establish prediction formulas of pulmonary function parameters and examined the effect of smoking on pulmonary function parameters. Methods We analyzed the result of parameters derived from the forced expiratory spirogram in 1,067 nonsmoking subjects from June in 1990 to December in 1991. They consisted of 306 males and 761 females and had neither respitatory symptoms nor history of respiratory disease. We derived prediction formulas by multiple linear regression method from their age, heights, and weights in each sex. To examine the effect of smoking on pulmonary function parameters, we classified 383 smoking men into three groups according to the past amount of smoking as follows : i.e. group of smokers who have smoked below 10 pack-years, 10-20 pack-years and above 20 pack-years. Regarding each group of past smoking as an independent dummy variable, we analyzed pulmonary function parameters including nonsmoking men as a baseline by multiple linear regression. We evaluated the smoking effect on pulmonary function parameters according to estimated p-value. Result : 1) Prediction formulas for pulmonary function parameters in each sex were derived. 2) The past smoking less than 10 pack-years does not give any effect on pulmonary function parameters. The past smoking of 10~20 pack-years showed significant negative correlation with $FEV_1$/FVC and FEF 25~75%, and the smoking above 20 pack years showed negative correlation with $FEV_1$ and $FEV_1$/FVC. Conclusion : We have got prediction formulas of pulmonary function parameters which is driven from forced expiratory spirogram in nonsmoking Korean adults by multiple linear regression from age, heights and weights of subjects. The past smoking more than 10 pack-years showed negative correlation with some pulmonary function parameters of airflow obstruction.