• 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.46 no.6
• /
• pp.817-825
• /
• 1999

Background : Forceps biopsy, bronchial brushing, and bronchial washing are used in conjunction with bronchoscopy to provide specimens for histologic and cytologic analysis in patients with suspected lung cancer. This study was performed to evaluate how many times brushing should be done and how much fluid should be used during bronchial washing for increasing diagnostic yield, and to evaluate which combination of these procedures gives the highest diagnostic yield. Methods : Forty patients, with suspected lung cancer, who had bronchoscopically visible lesions were enrolled in this prospective study. During one bronchoscopic examination four forceps biopsies, four bronchial brushings, and bronchial washing were done in all patients. The patients were divided into four groups by the amount of normal saline used for bronchial washing; group I, 10 ml ; group II, 20ml ; group III 30ml, and group IV, 40ml. We analyzed the results in 36 patients confirmed as lung cancer. Results : The diagnostic sensitivity of bronchial washing before and after forceps biopsy and bronchial brushing were 36% and 28%, respectively. The cumulative diagnostic sensitivity of bronchial washing was 47% and significantly higher than that of bronchial washing before or after forceps biopsy and bronchial brushing (p<0.05). The diagnostic sensitivity of bronchial washing with saline of 30ml was significantly higher than that of bronchial washing with saline of 10ml or 20ml (p<0.05). The diagnostic sensitivity of the first brushing was 75%, the second brushing 78%, the third brushing 83%, and the fourth brushing 67%. With repeated brushing up to three times, the diagnostic sensitivity increased to 92% (p<0.05). However, inclusion of the fourth brushing did not give a further increase of the diagnostic sensitivity. The diagnostic sensitivity of forceps biopsy was 86%. The diagnostic sensitivities of forceps biopsy by the type of bronchial lesion were as follows: tumor, 88%; infiltration, 67%; infiltration with nodularity, 80%; and collapse, 100%. The combination of forceps biopsy and bronchial washing gave a diagnostic sensitivity of 89%. The diagnostic sensitivity of combining forceps biopsy with bronchial brushing was 97%. Addition of bronchial washing did not increase the diagnostic yield over forceps biopsy and bronchial brushing. Conclusion : In patients with central lung cancer, forceps biopsies and repeated brushings up to three times should be done for maximal diagnostic yield.

• Journal of Intelligence and Information Systems
• /
• v.23 no.3
• /
• pp.45-67
• /
• 2017

According to the Internet Usage Research performed in 2016, the number of internet users and the internet usage have been increasing. Smartphone, compared to the computer, is taking a more dominant role as an internet access device. As the number of smart devices have been increasing, some views that the demand on high-speed internet will decrease; however, Despite the increase in smart devices, the high-speed Internet market is expected to slightly increase for a while due to the speedup of Giga Internet and the growth of the IoT market. As the broadband Internet market saturates, telecom operators are over-competing to win new customers, but if they know the cause of customer exit, it is expected to reduce marketing costs by more effective marketing. In this study, we analyzed the relationship between the cancellation rates of telecommunication products and the factors affecting them by combining the data of 3 cities, Anyang, Gunpo, and Uiwang owned by a telecommunication company with the regional data from KOSIS(Korean Statistical Information Service). Especially, we focused on the assumption that the neighboring areas affect the distribution of the cancellation rates by coupling type, so we conducted spatial cluster analysis on the 3 types of cancellation rates of each region using the spatial analysis tool, SatScan, and analyzed the various relationships between the cancellation rates and the regional data. In the analysis phase, we first summarized the characteristics of the clusters derived by combining spatial information and the cancellation data. Next, based on the results of the cluster analysis, Variance analysis, Correlation analysis, and regression analysis were used to analyze the relationship between the cancellation rates data and regional data. Based on the results of analysis, we proposed appropriate marketing methods according to the region. Unlike previous studies on regional characteristics analysis, In this study has academic differentiation in that it performs clustering based on spatial information so that the regions with similar cancellation types on adjacent regions. In addition, there have been few studies considering the regional characteristics in the previous study on the determinants of subscription to high-speed Internet services, In this study, we tried to analyze the relationship between the clusters and the regional characteristics data, assuming that there are different factors depending on the region. In this study, we tried to get more efficient marketing method considering the characteristics of each region in the new subscription and customer management in high-speed internet. As a result of analysis of variance, it was confirmed that there were significant differences in regional characteristics among the clusters, Correlation analysis shows that there is a stronger correlation the clusters than all region. and Regression analysis was used to analyze the relationship between the cancellation rate and the regional characteristics. As a result, we found that there is a difference in the cancellation rate depending on the regional characteristics, and it is possible to target differentiated marketing each region. As the biggest limitation of this study and it was difficult to obtain enough data to carry out the analyze. In particular, it is difficult to find the variables that represent the regional characteristics in the Dong unit. In other words, most of the data was disclosed to the city rather than the Dong unit, so it was limited to analyze it in detail. The data such as income, card usage information and telecommunications company policies or characteristics that could affect its cause are not available at that time. The most urgent part for a more sophisticated analysis is to obtain the Dong unit data for the regional characteristics. Direction of the next studies be target marketing based on the results. It is also meaningful to analyze the effect of marketing by comparing and analyzing the difference of results before and after target marketing. It is also effective to use clusters based on new subscription data as well as cancellation data.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.85-107
• /
• 2019

Online consumers browse products belonging to a particular product line or brand for purchase, or simply leave a wide range of navigation without making purchase. The research on the behavior and purchase of online consumers has been steadily progressed, and related services and applications based on behavior data of consumers have been developed in practice. In recent years, customization strategies and recommendation systems of consumers have been utilized due to the development of big data technology, and attempts are being made to optimize users' shopping experience. However, even in such an attempt, it is very unlikely that online consumers will actually be able to visit the website and switch to the purchase stage. This is because online consumers do not just visit the website to purchase products but use and browse the websites differently according to their shopping motives and purposes. Therefore, it is important to analyze various types of visits as well as visits to purchase, which is important for understanding the behaviors of online consumers. In this study, we explored the clustering analysis of session based on click stream data of e-commerce company in order to explain diversity and complexity of search behavior of online consumers and typified search behavior. For the analysis, we converted data points of more than 8 million pages units into visit units' sessions, resulting in a total of over 500,000 website visit sessions. For each visit session, 12 characteristics such as page view, duration, search diversity, and page type concentration were extracted for clustering analysis. Considering the size of the data set, we performed the analysis using the Mini-Batch K-means algorithm, which has advantages in terms of learning speed and efficiency while maintaining the clustering performance similar to that of the clustering algorithm K-means. The most optimized number of clusters was derived from four, and the differences in session unit characteristics and purchasing rates were identified for each cluster. The online consumer visits the website several times and learns about the product and decides the purchase. In order to analyze the purchasing process over several visits of the online consumer, we constructed the visiting sequence data of the consumer based on the navigation patterns in the web site derived clustering analysis. The visit sequence data includes a series of visiting sequences until one purchase is made, and the items constituting one sequence become cluster labels derived from the foregoing. We have separately established a sequence data for consumers who have made purchases and data on visits for consumers who have only explored products without making purchases during the same period of time. And then sequential pattern mining was applied to extract frequent patterns from each sequence data. The minimum support is set to 10%, and frequent patterns consist of a sequence of cluster labels. While there are common derived patterns in both sequence data, there are also frequent patterns derived only from one side of sequence data. We found that the consumers who made purchases through the comparative analysis of the extracted frequent patterns showed the visiting pattern to decide to purchase the product repeatedly while searching for the specific product. The implication of this study is that we analyze the search type of online consumers by using large - scale click stream data and analyze the patterns of them to explain the behavior of purchasing process with data-driven point. Most studies that typology of online consumers have focused on the characteristics of the type and what factors are key in distinguishing that type. In this study, we carried out an analysis to type the behavior of online consumers, and further analyzed what order the types could be organized into one another and become a series of search patterns. In addition, online retailers will be able to try to improve their purchasing conversion through marketing strategies and recommendations for various types of visit and will be able to evaluate the effect of the strategy through changes in consumers' visit patterns.

• Journal of The Korean Society of Grassland and Forage Science
• /
• v.4 no.1
• /
• pp.41-60
• /
• 1983

Experiments were conducted to study the effect of stage of maturity at harvest on the quality of silage. Herbage samples taken from the barley plant, rye plant, wheat plant, oat plant, Orchardgrass, Italian ryegrass, a mixed grass sward of Orchardgrass and Italian ryegrass and corn plant at different stages of maturity and ensiled in order to evaluate the effect of maturity on the chemical composition and feeding value as well as digestibility using sheep. Forage material were ensiled in small concrete silo. 1. The dry matter yield per 10a increased with advancing the maturity. Yield of brarley plant was 404, 635 and 900 kg at heading, milk and milk dough stage, respectively. Rye plant yield was 279, 589, 708, 10,000, 1,265, 1,376 and 1,492 kg at booting, before heading, early heading, late heading, early flowering, late flowering and after flowering stage, respectively. Italian ryegrass yield was 355, 613, 844 and 1,109 kg at vegetative, booting, heading and flowering, respectively. Orchardgrass/Italian ryegrass production was 477, 696, 891 and 1,027 kg at before was 458, 1,252, 1,534, 1,986 and 2,053 kg at tassel, early milk, yellow ripe and ripe stage, respectively. 2. Dry matter content increased with advancing maturity, but crude protein declined markedly. The NFE content decreased with advancing maturity of all the herbages except corn plant where NFE content increased, but corn plant increased. The content of crude fiber increased with advancing maturity except corn plant. The content of crude ash decreased with advancing maturity. In the rye plant, the content of neutral detergent fiber (NDF), acid detergent fiber (ADF) and cellulose increased with advancing maturity. 3. In vitro dry matter digestibilities of the rye plant was 53.6, 54.1, 50.7, 47.1, 44.9, 40.1 and 38.9% booting, before hcading, early heading, late heading, early flowering, late flowering and after flowering stage, respectively. The regression equation was $Y=56.22-0.74X+0.009X^2$ (X=cutting date from the first cut, Y=dry matter digestibilities). 4. In vitro digestible dry matter yield (kg/10a) of rye plant increased with advancing maturity, but declined from the flowering stage. The regression equation was $Y=168.88+26.09X-0.41X^2$ (X=cutting date from the first cut). 5. In vitro digestibility of dry matter in the corn plant was 69.2, 71.5, 69.8 and 69.9% at tassel, early milk, milk and yellow ripe stage, respectively. 6. The digestibility of crude protein and crude fiber of all plants decreased with advancing matuity, but NFE of the barley and corn generally increased. 7. The TDN contents on the dry matter basis decreased, but those of barley and corn silage were not different. TDN content of barley was 57.8, 57.1 and 57.9% at heading, milk and milk dough stage, respectively. That of rye silage was 50.0, 27.2 and 43.7% at early flowering, after flowering and milk stage, respectively. Italian ryegrass silage was 67.9, 63.7, and 54.9% at before heading, early heading and after heading, respectively. In case of Orchardgrass silage the TDN was 54.8, 52.9 and 46.1% at after heading, after flowering and milk, respectively. Corn shows TDN value of 59.5, 62.8 and 61.6% at milk, yellow ripe and ripe, respectively. 8. The pH value increased slightly by advancing maturity. 9. the content of organic acid decreased by advancing maturity and also increasing the DM content.

• Tuberculosis and Respiratory Diseases
• /
• v.43 no.6
• /
• pp.987-996
• /
• 1996

Background : We prospectively evaluated the applicability and effect of noninvasive ventilation (NIV) in acute respiratory failure and tried to find out the parameters that could predict successful application of NIV. Methods : Twenty-six out of 106 patients with either acute ventilatory failure (VF: $PaCO_2$ > 43 mm Hg with pH < 7.35) or oxygenation failure (OF: $PaO_2/AO_2$ < 300 mm Hg with $pH{\geq}7.35$) requiring mechanical ventilation were managed by NIV (CPAP + pressure suppon, or BiPAP) with face mask. Eleven out of 19 cases with VF (57.9%) (M : F=7 : $55.4{\pm}14.6$ yrs) and 15 out of 87 cases with OF (17.2%) (M : F=12 : 3, $50.6{\pm}15.6$ yrs) were s uilable for NIY. Respiratory rates, arterial blood gases and success rate of NIV were analyzed in each group. Results: 81.8% (9/11) of YF and 40% (6/15) of OF were successfully managed on NIV and were weruled from mechanical ventilator without resorting to endotracheal intubation. Complications were noted in 2 cases (nasal skin necrosis 1, gaseous gastric distension 1). In NIV for ventilatory failure, the respiration rate was significantly decreased at 12 hour of NIV ($34{\pm}9$ /min pre-NIV, $26{\pm}6$ /min at 12 hour of NIV, p=0.045), while $PaCO_2$ ($87.3{\pm}20.6$ mm Hg pre-NIV, $81.2{\pm}9.1$ mm Hg at 24 hour of NIV) and pH ($7.26{\pm}0.04$, $7.32{\pm}0.02$, respectively, p <0.05) were both significantly decreased at 24 hour of NIV In NIV for oxygenation failure, $PaCO_2$ were not different between the successful and the failed cases at pre-NIV and till 12 hours after NIV. The $PaO_2/FIO_2$ ratio, however, significantly improved at 0.5 hour of NIV in successful cases and were maintained at around 200 mm Hg (n=6 : at baseline, 0.5h, 6h, 12h : $120.0{\pm}19.6$, $218.9{\pm}98.3$, $191.3{\pm}55.2$, $232.8{\pm}17.6$ mm Hg, respectively, p=0.0211), but it did not rise in the failed cases (n=9 : $127.9{\pm}63.0$, $116.8{\pm}24.4$, $100.6{\pm}34.6$, $129.8{\pm}50.3$ mm Hg, respectively, p=0.5319). Conclusion : From the above results we conclude that NIV is effective for hypercapnic respiratory failure and its success was heralded by reduction of respiration rale before the reduction in $PaCO_2$ level. In hypoxic respiratory failure, NIV is much less effective, and the immediate improvement of $PaO_2/FIO_2$ ratio at 0.5h after application is thought to be a predictor of successful NIV.