• The Korean Journal of Nuclear Medicine
• /
• v.38 no.4
• /
• pp.318-324
• /
• 2004

Purpose: Philips GEMINI is a newly introduced whole-body GSO PET/CT scanner. In this study, performance of the scanner including spatial resolution, sensitivity, scatter fraction, noise equivalent count ratio (NECR) was measured utilizing NEMA NU2-2001 standard protocol and compared with performance of LSO, BGO crystal scanner. Methods: GEMINI is composed of the Philips ALLEGRO PET and MX8000 D multi-slice CT scanners. The PET scanner has 28 detector segments which have an array of 29 by 22 GSO crystals ($4{\times}6{\times}20$ mm), covering axial FOV of 18 cm. PET data to measure spatial resolution, sensitivity, scatter fraction, and NECR were acquired in 3D mode according to the NEMA NU2 protocols (coincidence window: 8 ns, energy window: $409[\sim}664$ keV). For the measurement of spatial resolution, images were reconstructed with FBP using ramp filter and an iterative reconstruction algorithm, 3D RAMLA. Data for sensitivity measurement were acquired using NEMA sensitivity phantom filled with F-18 solution and surrounded by $1{\sim}5$ aluminum sleeves after we confirmed that dead time loss did not exceed 1%. To measure NECR and scatter fraction, 1110 MBq of F-18 solution was injected into a NEMA scatter phantom with a length of 70 cm and dynamic scan with 20-min frame duration was acquired for 7 half-lives. Oblique sinograms were collapsed into transaxial slices using single slice rebinning method, and true to background (scatter+random) ratio for each slice and frame was estimated. Scatter fraction was determined by averaging the true to background ratio of last 3 frames in which the dead time loss was below 1%. Results: Transverse and axial resolutions at 1cm radius were (1) 5.3 and 6.5 mm (FBP), (2) 5.1 and 5.9 mm (3D RAMLA). Transverse radial, transverse tangential, and axial resolution at 10 cm were (1) 5.7, 5.7, and 7.0 mm (FBP), (2) 5.4, 5.4, and 6.4 mm (3D RAMLA). Attenuation free values of sensitivity were 3,620 counts/sec/MBq at the center of transaxial FOV and 4,324 counts/sec/MBq at 10 cm offset from the center. Scatter fraction was 40.6%, and peak true count rate and NECR were 88.9 kcps @ 12.9 kBq/mL and 34.3 kcps @ 8.84 kBq/mL. These characteristics are better than that of ECAT EXACT PET scanner with BGO crystal. Conclusion: The results of this field test demonstrate high resolution, sensitivity and count rate performance of the 3D PET/CT scanner with GSO crystal. The data provided here will be useful for the comparative study with other 3D PET/CT scanners using BGO or LSO crystals.

• Journal of Preventive Medicine and Public Health
• /
• v.22 no.1 s.25
• /
• pp.29-44
• /
• 1989

A household survey was conducted to determine the utilization rate of medical facilities and to identify the factors related with the utilization in the South District of Taegu from July 3 to July 15, 1988. Study population included 1,723 family members of 431 households which were selected by one-stage simple cluster random sampling. Well trained medical college students interviewed mainly housewives with a structurized questionnaire. Morbidity rate of acute illness during the 2-week period was 101 per 1,000 persons and it was highest in the age group of 9 years below. The rate for chronic illness was 77 per 1,000 persons, increasing with age, low income and medicaid benefit. During the 2-week period, 689 of 1,000 persons utilized the medical facilities. Of the facilities, most number, 294, used hospital and clinic, and the order ran as pharmacy, health center, and herb medical clinic. The utilization rate was higher in the female, 70-year and older group, medicaid group, the lowest income class and self-employed group than other groups. The average number of visits among users of medical facilities during the 2-week period was 3.25. those who visited medical facilities most frequently were females, the 70-year and older group, the lowest income class and blue collar worker group. During one-year period, admission rate of 1,000 persons was 27.6 and that of female was 38.9, higher than that of male. the eldest group had the highest admission rate. Admission rate of medical insurance beneficiaries was twice or higher than non-beneficiaries. The higher the family monthly income, the more frequently they admitted. During one-year period, average admission days of the persons hospitalized were 22.5 days and males were hospitalized longer than females. The groups which were hospitalized longest were those between the ages of 40 and 49, medical insurance beneficiaries, the lowest income group and unemployed group. During one-year period, average admission days of 1,000 persons were 560 days and those of female were 661 days, more than those of male. The guoups which had the longest admission days were those above 70 years of age, the lowest income and unemployed groups. The medical insurance beneficiaries were three times or longer than non-beneficiaries. In logistic regression analysis of utilization of physician significant independent variables were the 9-year and younger group(+), the 70-year and older group(+), acute illness episode(+), chronic illness episode(+), medical insurance beneficiary(+) and white collar workers(-). Acute and chronic illness episode(+), and medical insurance for government employees and private school teacher(-) were significant variables in analysis of utilization of pharmacy. In multiple regression analysis of the number of physician visits, siginificant variables were acute illnes episode(+), chronic illness episode(+), industrial, occupational and regional medical insurance beneficiary(+), white collar workers(-). Acute and chronic illness episode(+), and medical insurance beneficiary(-) were significant variables in analysis of the number of pharmacy visits. In logistic regression analysis of admission event, significant independent variables were the 9-year and younger group(+), the 70-year and older group(+) , chronic illness episode(+), and medical insurance beneficiary(+).

• Journal of Preventive Medicine and Public Health
• /
• v.27 no.2 s.46
• /
• pp.258-273
• /
• 1994

To compare the physical complaints of vinylhouse farmers with those of non-vinylhouse farmers, a personal interviews on 250 vinylhouse and 142 non-vinylhouse farmers were conducted in Sungjoo county in Kyungpook province selected by a random sampling from July 5 to July 10, 1993. Blood pressure of the subjects was also measured. Vinylhouse farmers had a higher average age, larger family size, shorter experience of farming, more working hours per day and working days per year and higher annual income than the non-vinylhouse farmers. The frequency of pesticide spray of the vinylhouse farmers was 3.4 times on the average in June 1993 as compared with 2.0 times of non-vinylhouse farmers, and 16.7 times for the vinylhouse farmers during the last one year while it was 8.3 times for the non-vinylhouse farmers in the same period. While 39.6% of vinylhouse farmers experienced pesticide intoxication symptoms such as headache, nausea, vomiting, dizziness, itching, and skin irritation, etc. during the month of June, 25.4% of non-vinylhouse farmers experienced such symptoms. The most frequent symptoms among eight symptoms that constitute the farmers' syndrome were lumbago, numbness of hand or foot, shoulder pain and dizziness regardless of sex and type of farming. Prevalence of the farmers' syndrome in male and female among vinylhouse farmers were 22.1%, 43.4%, respectively, and the prevalence in non-vinylhouse farmers was 23.2% for male and 50.7% for female. There was no statistically significant difference in the prevalence of farmers' syndrome between vinylhouse and non-vinylhouse farmers. However, the prevalence in female was about 2 times higher than that of male. When the effects of other factors were adjusted by multiple logistic regression for farmers' syndrome, the prevalence in female was 3.0 times higher than that of male. The prevalence of farmers' syndrome was increased as the age of farmers increased in both vinylhouse and non-vinylhouse farmers, and adjusted odds ratio of farmers' syndrome increased by 3% as the age increased by 1 year. Adjusted odds ratio for Farmers' syndrome in farmers who experienced pesticide intoxication during the month of June was 3.1 times higher than that of farmers who did not have such experience. While the prevalence of hypertension in male and female non-vinylhouse farmers were 22.4%, 13.7%, respectively, the prevalence in vinylhouse farmers were 13.5% for male and 12.0% for female. However, there was no association between farmers' syndrome and hypertension. It was found in this study that the vinylhouse farmers are at a high risk of pesticide intoxication, which is associated with tile common physical complaints. To reduce such risk it is necessary to develop farming methods which do not require the pesticide or may use less pesticide, a safer method of pesticide spraying, and the protective equipments which can be worn at a high temperature and have a better protective effect. Also education of farmers for the correct methods of ventilation after pesticide spraying in the vinylhouse and wearing the protective equipments may be considered as a supportive method. Since inappropriate posture at work and intensive labor may cause farmers' syndrome, it is recommended to develop farming tools which reduce physical burden and take a rest and exercise periodically during work. It is necessary to strengthen the hypertension management program of the Kyungpook province, because the prevalence of hypertension was as high as about 15%.

• Radiation Oncology Journal
• /
• v.23 no.1
• /
• pp.22-31
• /
• 2005

Purpose : To conduct a nationwide survey on the principals in radiotherapy for rectal cancer, and produce a database of Korean Patterns of Care Study. Materials and Methods : We developed web-based Patterns of Care Study system and a national survey was conducted using random sampling based on power allocation methods. Eligible patients were who had postoperative radiotherapy for rectal cancer without gross residual tumor after surgical resection and without previous history of other cancer and radiotherapy to pelvis. Data of patients were Inputted to the web based PCS system by each investigators in 19 institutions. Results : Informations on 309 patients with rectal cancer who received radiotherapy between 1998 and 1999 were collected. Male to female ratio was 59 : 41, and the most common location of tumor was lower rectum ($46\%$). Preoperative CEA was checked in $79\%$ of cases and its value was higher than 6 ng/ml in $32\%$. Pathologic stage were I in $1.5\%$, II in $32\%$, III in $53\%$, and IV in $1.6\%$. Low anterior resection was the most common type of surgery and complete resection was peformed in $95\%$ of cases. Distal resection margin was less than 2 cm in $30\%$, and number of lymph node dissected was less than 12 in $31\%$. Chemotherapy was peformed in $91\%$ and most common regimen was 5-FU and leucovorine ($59\%$). The most common type of field arrangement used for the initial pelvic field was the four field box (Posterior-Right-Left) technique ($65.0\%$), and there was no AP-PA parallel opposing field used. Patient position was prone in $81.2\%$, and the boost field was used in $61.8\%$. To displace bowel outward, pressure modulating devices or bladder filling was used in $40.1\%$. Radiation dose was prescribed to isocenter in $45.3\%$ and to isodose line in 123 cases ($39.8\%$). Percent delivered dose over $90\%$ was achieved in $92.9\%$. Conclusion : We could find the Patterns of Care for the radiotherapy in Korean rectal cancer patients was similar to that of US national survey. The type of surgery and the regimen of chemotherapy were variable according to institutions and the variations of radiation dose and field arrangement were within acceptable range.

• Journal of Global Scholars of Marketing Science
• /
• v.20 no.3
• /
• pp.269-277
• /
• 2010

Previous studies have shown that the most important factor affecting customer loyalty in the service industry is service quality. However, on the subject of whether service quality has a direct or indirect effect on customer loyalty, scholars' views apparently vary. Some studies suggest that service quality has a direct and fundamental influence on customer loyalty (Bai and Liu, 2002). However, others have shown that service quality not only directly affects customer loyalty, it also has an indirect impact on customer loyalty by influencing customer satisfaction and perceived value (Cronin, Brady, and Hult, 2000). Currently, there are few domestic articles that specifically address the relationship between service quality and customer loyalty in the mobile communication industry. Moreover, research has studied customer loyalty as a whole variable, rather than breaking it down further into multiple dimensions. Based on this analysis, this paper summarizes previous study results, establishes an effect mechanism model among service quality, customer satisfaction, and customer loyalty in the mobile communication industry, and presents a statistical test on model assumptions by using customer investigation data from Heilongjiang Mobile Company. It provides theoretical guidance for mobile service management based on the discussion of the hypothesis test results. For data collection, the sample comprised mobile users in Harbin city, and the survey was taken by random sampling. Out of a total of 300 questionnaires, 276 (92.9%) were recovered. After excluding invalid questionnaires, 249 remained, for an effective rate of 82.6 percent for the study. Cronbach's ${\alpha}$ coefficient was adapted to assess the scale reliability, and validity testing was conducted on the questionnaire from three aspects: content validity, construct validity. and convergent validity. The study tested for goodness of fit mainly from the absolute and relative fit indexes. From the hypothesis testing results, overall, four assumptions have not been supported. The ultimate affective relationship of service quality, customer satisfaction, and customer loyalty is demonstrated in Figure 2. On the whole, the service quality of the communication industry not only has a direct positive significant effect on customer loyalty, it also has an indirect positive significant effect on customer loyalty through service quality; the affective mechanism and extent of customer loyalty are different, and are influenced by each dimension of service quality. This study used the questionnaires of existing literature from home and abroad and tested them in empirical research, with all questions adapted to seven-point Likert scales. With the SERVQUAL scale of Parasuraman, Zeithaml, and Berry (1988), or PZB, as a reference point, service quality was divided into five dimensions-tangibility, reliability, responsiveness, assurance, and empathy-and the questions were simplified down to nineteen. The measurement of customer satisfaction was based mainly on Fornell (1992) and Wang and Han (2003), ending up with four questions. Based on the study’s three indicators of price tolerance, first choice, and complaint reaction were used to measure attitudinal loyalty, while repurchase intention, recommendation, and reputation measured behavioral loyalty. The collection and collation of literature data produced a model of the relationship among service quality, customer satisfaction, and customer loyalty in mobile communications, and China Mobile in the city of Harbin in Heilongjiang province was used for conducting an empirical test of the model and obtaining some useful conclusions. First, service quality in mobile communication is formed by the five factors mentioned earlier: tangibility, reliability, responsiveness, assurance, and empathy. On the basis of PZB SERVQUAL, the study designed a measurement scale of service quality for the mobile communications industry, and obtained these five factors through exploratory factor analysis. The factors fit basically with the five elements, indicating the concept of five elements of service quality for the mobile communications industry. Second, service quality in mobile communications has both direct and indirect positive effects on attitudinal loyalty, with the indirect effect being produced through the intermediary variable, customer satisfaction. There are also both direct and indirect positive effects on behavioral loyalty, with the indirect effect produced through two intermediary variables: customer satisfaction and attitudinal loyalty. This shows that better service quality and higher customer satisfaction will activate the attitudinal to service providers more active and show loyalty to service providers much easier. In addition, the effect mechanism of all dimensions of service quality on all dimensions of customer loyalty is different. Third, customer satisfaction plays a significant intermediary role among service quality and attitudinal and behavioral loyalty, indicating that improving service quality can boost customer satisfaction and make it easier for satisfied customers to become loyal customers. Moreover, attitudinal loyalty plays a significant intermediary role between service quality and behavioral loyalty, indicating that only attitudinally and behaviorally loyal customers are truly loyal customers. The research conclusions have some indications for Chinese telecom operators and others to upgrade their service quality. Two limitations to the study are also mentioned. First, all data were collected in the Heilongjiang area, so there might be a common method bias that skews the results. Second, the discussion addresses the relationship between service quality and customer loyalty, setting customer satisfaction as mediator, but does not consider other factors, like customer value and consumer features, This research will be continued in the future.

• Tuberculosis and Respiratory Diseases
• /
• v.44 no.3
• /
• pp.592-600
• /
• 1997

Background : The level of imposed work of breathing (WOB) is important for patient-ventilator synchrony and during weaning from mechanical ventilation. Triggering methods and the sensitivity of demand system are important determining factors of the imposed WOB. Flow triggering method is available on several modern ventilator and is believed to impose less work to a patient-triggered breath than pressure triggering method. We intended to compare the level of imposed WOB on two different methods of triggering and also at different levels of sensitivities on each triggering method (0.7 L/min vs 2.0 L/min on flow triggering ; $-1\;cmH_2O$ vs $-2cm\;H_2O$ on pressure triggering). Methods : The subjects were 12 patients ($64.8{\pm}4.2\;yrs$) on mechanical ventilation and were stable in respiratory pattern on CPAP $3\;cmH_2O$. Four different triggering sensitivities were applied at random order. For determination of imposed WOB, tracheal end pressure was measured through the monitoring lumen of Hi-Lo Jet tracheal tube (Mallincrodt, New York, USA) using pneumotachograph/pressure transducer (CP-100 pulmonary monitor, Bicore, Irvine, CA, USA). Other data of respiratory mechanics were also obtained by CP-100 pulmonary monitor. Results : The imposed WOB was decreased by 37.5% during 0.7 L/min on flow triggering compared to $-2\;cmH_2O$ on pressure triggering and also decreased by 14% during $-1\;cmH_2O$ compared to $-2\;cmH_2O$ on pressure triggering (p < 0.05 in each). The PTP(Pressure Time Product) was also decreased significantly during 0.7 L/min on flow triggering and $-1\;cmH_2O$ on pressure triggering compared to $-2\;cmH_2O$ on pressure triggering (p < 0.05 in each). The proportions of imposed WOB in total WOB were ranged from 37% to 85% and no significant changes among different methods and sensitivities. The physiologic WOB showed no significant changes among different triggering methods and sensitivities. Conclusion : To reduce the imposed WOB, flow triggering with sensitivity of 0.7 L/min would be better method than pressure triggering with sensitivity of $-2\;cm\;H_2O$.

• Tuberculosis and Respiratory Diseases
• /
• v.44 no.6
• /
• pp.1318-1325
• /
• 1997

Background : Since the late 1960s, mechanical ventilation has been accomplished primarily using volume controlled ventilation(VCV). While VCV allows a set tidal volume to be guaranteed, VCV could bring about excessive airway pressures that may be lead to barotrauma in the patients with acute lung injury. With the increment of knowledge related to ventilator-induced lung injury, pressure controlled ventilation(PCV) has been frequently applied to these patients. But, PCV has a disadvantage of variable tidal volume delivery as pulmonary impedance changes. Since the concept of combining the positive attributes of VCV and PCV(dual control ventilation, DCV) was described firstly in 1992, a few DCV modes were introduced. Pressure-regulated volume control(PRVC) mode, a kind of DCV, is pressure-limited, time-cycled ventilation that uses tidal volume as a feedback control for continuously adjusting the pressure limit However, no clinical studies were published on the efficacy of PRVC until now. 'This investigation studied the efficacy of PRVC in the patients with unstable respiratory mechanics. Methods : The subjects were 8 mechanically ventilated patients(M : F=6 : 2, $56{\pm}26$ years) who showed unstable respiratory mechanics, which was defined by the coefficients of variation of peak inspiratory pressure for 15 minutes greater than 10% under VCV, or the coefficients of variation of tidal volume greater than 10% under PCV. The study was consisited of 3 modes application with VCV, PCV and PRVC for 15 minutes by random order. To obtain same tidal volume, inspiratory pressure setting was adjusted in PCV. Respiratory parameters were measured by pulmonary monitor(CP-100 pulmonary monitor, Bicore, Irvine, CA, USA). Results : 1) Mean tidal volumes($V_T$) in each mode were not different(VCV, $431{\pm}102ml$ ; PCV, $417{\pm}99ml$ ; PRVC, $414{\pm}97ml$) 2) The coefficient of variation(CV) of $V_T$ were $5.2{\pm}3.9%$ in VCV, $15.2{\pm}7.5%$ in PCV and $19.3{\pm}10.0%$ in PRVC. The CV of $V_T$ in PCV and PRVC were significantly greater than that in VCV(p<0.01). 3) Mean peak inspiratory pressure(PIP) in VCV($31.0{\pm}6.9cm$ $H_2O$) was higher than PIP in PCV($26.0{\pm}6.5cm$ $H_2O$) or PRVC($27.0{\pm}6.4cm$ $H_2O$)(p<0.05). 4) The CV of PIP were $13.9{\pm}3.7%$ in VCV, $4.9{\pm}2.6%$ in PVC and $12.2{\pm}7.0%$ in PRVC. The CV of PIP in VCV and PRVC were greater than that in PCV(p<0.01). Conclusions : Because of wide fluctuations of VT and PIP, PRVC mode did not seem to have advantages compared to VCV or PCV in the patients with unstable respiratory mechanics.

• Journal of Intelligence and Information Systems
• /
• v.26 no.3
• /
• pp.149-169
• /
• 2020

"The Urban Renewal New Deal project", one of the government's major national projects, is about developing underdeveloped areas by investing 50 trillion won in 100 locations on the first year and 500 over the next four years. This project is drawing keen attention from the media and local governments. However, the project model which fails to reflect the original characteristics of the area as it divides project area into five categories: "Our Neighborhood Restoration, Housing Maintenance Support Type, General Neighborhood Type, Central Urban Type, and Economic Base Type," According to keywords for successful urban regeneration in Korea, "resident participation," "regional specialization," "ministerial cooperation" and "public-private cooperation", when local governments propose urban regeneration projects to the government, they can see that it is most important to accurately understand the characteristics of the city and push ahead with the projects in a way that suits the characteristics of the city with the help of local residents and private companies. In addition, considering the gentrification problem, which is one of the side effects of urban regeneration projects, it is important to select and implement urban regeneration types suitable for the characteristics of the area. In order to supplement the limitations of the 'Urban Regeneration New Deal Project' methodology, this study aims to propose a system that recommends urban regeneration types suitable for urban regeneration sites by utilizing various machine learning algorithms, referring to the urban regeneration types of the '2025 Seoul Metropolitan Government Urban Regeneration Strategy Plan' promoted based on regional characteristics. There are four types of urban regeneration in Seoul: "Low-use Low-Level Development, Abandonment, Deteriorated Housing, and Specialization of Historical and Cultural Resources" (Shon and Park, 2017). In order to identify regional characteristics, approximately 100,000 text data were collected for 22 regions where the project was carried out for a total of four types of urban regeneration. Using the collected data, we drew key keywords for each region according to the type of urban regeneration and conducted topic modeling to explore whether there were differences between types. As a result, it was confirmed that a number of topics related to real estate and economy appeared in old residential areas, and in the case of declining and underdeveloped areas, topics reflecting the characteristics of areas where industrial activities were active in the past appeared. In the case of the historical and cultural resource area, since it is an area that contains traces of the past, many keywords related to the government appeared. Therefore, it was possible to confirm political topics and cultural topics resulting from various events. Finally, in the case of low-use and under-developed areas, many topics on real estate and accessibility are emerging, so accessibility is good. It mainly had the characteristics of a region where development is planned or is likely to be developed. Furthermore, a model was implemented that proposes urban regeneration types tailored to regional characteristics for regions other than Seoul. Machine learning technology was used to implement the model, and training data and test data were randomly extracted at an 8:2 ratio and used. In order to compare the performance between various models, the input variables are set in two ways: Count Vector and TF-IDF Vector, and as Classifier, there are 5 types of SVM (Support Vector Machine), Decision Tree, Random Forest, Logistic Regression, and Gradient Boosting. By applying it, performance comparison for a total of 10 models was conducted. The model with the highest performance was the Gradient Boosting method using TF-IDF Vector input data, and the accuracy was 97%. Therefore, the recommendation system proposed in this study is expected to recommend urban regeneration types based on the regional characteristics of new business sites in the process of carrying out urban regeneration projects."

• Journal of Intelligence and Information Systems
• /
• v.18 no.2
• /
• pp.29-45
• /
• 2012

Bond rating is regarded as an important event for measuring financial risk of companies and for determining the investment returns of investors. As a result, it has been a popular research topic for researchers to predict companies' credit ratings by applying statistical and machine learning techniques. The statistical techniques, including multiple regression, multiple discriminant analysis (MDA), logistic models (LOGIT), and probit analysis, have been traditionally used in bond rating. However, one major drawback is that it should be based on strict assumptions. Such strict assumptions include linearity, normality, independence among predictor variables and pre-existing functional forms relating the criterion variablesand the predictor variables. Those strict assumptions of traditional statistics have limited their application to the real world. Machine learning techniques also used in bond rating prediction models include decision trees (DT), neural networks (NN), and Support Vector Machine (SVM). Especially, SVM is recognized as a new and promising classification and regression analysis method. SVM learns a separating hyperplane that can maximize the margin between two categories. SVM is simple enough to be analyzed mathematical, and leads to high performance in practical applications. SVM implements the structuralrisk minimization principle and searches to minimize an upper bound of the generalization error. In addition, the solution of SVM may be a global optimum and thus, overfitting is unlikely to occur with SVM. In addition, SVM does not require too many data sample for training since it builds prediction models by only using some representative sample near the boundaries called support vectors. A number of experimental researches have indicated that SVM has been successfully applied in a variety of pattern recognition fields. However, there are three major drawbacks that can be potential causes for degrading SVM's performance. First, SVM is originally proposed for solving binary-class classification problems. Methods for combining SVMs for multi-class classification such as One-Against-One, One-Against-All have been proposed, but they do not improve the performance in multi-class classification problem as much as SVM for binary-class classification. Second, approximation algorithms (e.g. decomposition methods, sequential minimal optimization algorithm) could be used for effective multi-class computation to reduce computation time, but it could deteriorate classification performance. Third, the difficulty in multi-class prediction problems is in data imbalance problem that can occur when the number of instances in one class greatly outnumbers the number of instances in the other class. Such data sets often cause a default classifier to be built due to skewed boundary and thus the reduction in the classification accuracy of such a classifier. SVM ensemble learning is one of machine learning methods to cope with the above drawbacks. Ensemble learning is a method for improving the performance of classification and prediction algorithms. AdaBoost is one of the widely used ensemble learning techniques. It constructs a composite classifier by sequentially training classifiers while increasing weight on the misclassified observations through iterations. The observations that are incorrectly predicted by previous classifiers are chosen more often than examples that are correctly predicted. Thus Boosting attempts to produce new classifiers that are better able to predict examples for which the current ensemble's performance is poor. In this way, it can reinforce the training of the misclassified observations of the minority class. This paper proposes a multiclass Geometric Mean-based Boosting (MGM-Boost) to resolve multiclass prediction problem. Since MGM-Boost introduces the notion of geometric mean into AdaBoost, it can perform learning process considering the geometric mean-based accuracy and errors of multiclass. This study applies MGM-Boost to the real-world bond rating case for Korean companies to examine the feasibility of MGM-Boost. 10-fold cross validations for threetimes with different random seeds are performed in order to ensure that the comparison among three different classifiers does not happen by chance. For each of 10-fold cross validation, the entire data set is first partitioned into tenequal-sized sets, and then each set is in turn used as the test set while the classifier trains on the other nine sets. That is, cross-validated folds have been tested independently of each algorithm. Through these steps, we have obtained the results for classifiers on each of the 30 experiments. In the comparison of arithmetic mean-based prediction accuracy between individual classifiers, MGM-Boost (52.95%) shows higher prediction accuracy than both AdaBoost (51.69%) and SVM (49.47%). MGM-Boost (28.12%) also shows the higher prediction accuracy than AdaBoost (24.65%) and SVM (15.42%)in terms of geometric mean-based prediction accuracy. T-test is used to examine whether the performance of each classifiers for 30 folds is significantly different. The results indicate that performance of MGM-Boost is significantly different from AdaBoost and SVM classifiers at 1% level. These results mean that MGM-Boost can provide robust and stable solutions to multi-classproblems such as bond rating.

• Journal of Intelligence and Information Systems
• /
• v.25 no.4
• /
• pp.141-154
• /
• 2019

Rapid growth of internet technology and social media is progressing. Data mining technology has evolved to enable unstructured document representations in a variety of applications. Sentiment analysis is an important technology that can distinguish poor or high-quality content through text data of products, and it has proliferated during text mining. Sentiment analysis mainly analyzes people's opinions in text data by assigning predefined data categories as positive and negative. This has been studied in various directions in terms of accuracy from simple rule-based to dictionary-based approaches using predefined labels. In fact, sentiment analysis is one of the most active researches in natural language processing and is widely studied in text mining. When real online reviews aren't available for others, it's not only easy to openly collect information, but it also affects your business. In marketing, real-world information from customers is gathered on websites, not surveys. Depending on whether the website's posts are positive or negative, the customer response is reflected in the sales and tries to identify the information. However, many reviews on a website are not always good, and difficult to identify. The earlier studies in this research area used the reviews data of the Amazon.com shopping mal, but the research data used in the recent studies uses the data for stock market trends, blogs, news articles, weather forecasts, IMDB, and facebook etc. However, the lack of accuracy is recognized because sentiment calculations are changed according to the subject, paragraph, sentiment lexicon direction, and sentence strength. This study aims to classify the polarity analysis of sentiment analysis into positive and negative categories and increase the prediction accuracy of the polarity analysis using the pretrained IMDB review data set. First, the text classification algorithm related to sentiment analysis adopts the popular machine learning algorithms such as NB (naive bayes), SVM (support vector machines), XGboost, RF (random forests), and Gradient Boost as comparative models. Second, deep learning has demonstrated discriminative features that can extract complex features of data. Representative algorithms are CNN (convolution neural networks), RNN (recurrent neural networks), LSTM (long-short term memory). CNN can be used similarly to BoW when processing a sentence in vector format, but does not consider sequential data attributes. RNN can handle well in order because it takes into account the time information of the data, but there is a long-term dependency on memory. To solve the problem of long-term dependence, LSTM is used. For the comparison, CNN and LSTM were chosen as simple deep learning models. In addition to classical machine learning algorithms, CNN, LSTM, and the integrated models were analyzed. Although there are many parameters for the algorithms, we examined the relationship between numerical value and precision to find the optimal combination. And, we tried to figure out how the models work well for sentiment analysis and how these models work. This study proposes integrated CNN and LSTM algorithms to extract the positive and negative features of text analysis. The reasons for mixing these two algorithms are as follows. CNN can extract features for the classification automatically by applying convolution layer and massively parallel processing. LSTM is not capable of highly parallel processing. Like faucets, the LSTM has input, output, and forget gates that can be moved and controlled at a desired time. These gates have the advantage of placing memory blocks on hidden nodes. The memory block of the LSTM may not store all the data, but it can solve the CNN's long-term dependency problem. Furthermore, when LSTM is used in CNN's pooling layer, it has an end-to-end structure, so that spatial and temporal features can be designed simultaneously. In combination with CNN-LSTM, 90.33% accuracy was measured. This is slower than CNN, but faster than LSTM. The presented model was more accurate than other models. In addition, each word embedding layer can be improved when training the kernel step by step. CNN-LSTM can improve the weakness of each model, and there is an advantage of improving the learning by layer using the end-to-end structure of LSTM. Based on these reasons, this study tries to enhance the classification accuracy of movie reviews using the integrated CNN-LSTM model.