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

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

• Journal of Mushroom
• /
• v.2 no.3
• /
• pp.115-126
• /
• 2004

Somatic hybrids of inter-compatible and inter-incompatible strains were obtained by protoplast fusion. The fusion products between compatible strains, Pleurotus ostreatus and P. florida, formed heterokaryons, while fusants between incompatible strains such as P. cornucopiae + P. florida, P. ostreatus + Ganoderma applanatum, P. florida + Ganoderma lucidum, and P. ostreatus + Flammulina velutipes formed synkaryons that retained genes from both parents. The heterokaryons showed the same level of basidioma development. In contrast, the synkaryons showed unique characteristics including clamp connection formation at mitosis, either partner basidioma development, and abnormal segregation and recombination compared with inter-compatible strains. Synkaryons can be classified into homokaryoyic and heterokaryotic type. A comparison of somatic hybrids with compatible and incompatible strains was made using random amplified polymorphic DNA (RAPD) analysis. The heterokaryons between compatible species showed the same level of variability and contained both parental RAPD bands. In contrast, most of the synkaryons between incompatible species showed similarity to those of either parental bands and non-parental RAPD bands. Synkaryons can be classified into microgenome insertion type and macrogenome insertion type. A tetrapolar mating system was found among monospore isolates in somatic hybrids and wild type P. ostreatus. Homokaryons from each somatic hybrid combination were paired with tester homokaryons of the initial wild type of P. ostreatus. The changed mating types were identified in progenies. The pattern of mating type switching in somatic hybrids depends on compatibility of fusion partner. There are several factors related to the mechanism of clamp connection formation and fruiting body development of synkaryons. Of these,the major factor may be associated with self-fertility and mating type switching such as homokaryotic fruiting of wild type P. ostreatus. This review will discuss these aspects.

• Journal of agricultural medicine and community health
• /
• v.8 no.1
• /
• pp.44-65
• /
• 1983

As a part of the epidemiological studies of clonorchiasis, this study was conducted to evaluate the current endemicity and the natural transition of the Clonorchis infection in Gimhae Gun and delta area a high endemic area in Korea in recent years, prior to the introduction of praziquantel which will eventually influence the status of the prevalence. The data obtained in this study in 1983 were evaluated for natural transition of the infection in comparison with those obtained 16 years ago in 1967 by the author(Kim, 1974). The areas of investigation, villages and schools surveyed, methods and techniques used in this study were the same as in 1967, except for the contents of the questionnaire for raw freshwater fish consumption by the local inhabitants. 1) The prevalence rate of clonorchiasis in the general population of the villages was 48.1% on the average out of a total of 484 persons examined. The average of those of the riverside-delta area was 65.2% and 43.0% in the inland area. Among the schoolchildren, the prevalence rate was 8.2% on the average out of a total of 1,423 examined. By area, the prevalence rate was 10.8% in the riverside delta area and 2.8% in the inland area. By sex, difference in the prevalence was seen only in the inhabitants of the inland area showing 52.4% in the male and 33.5% in the female. 2) In the natural transition of the infection, the prevalence rate among the inhabitants has decreased from 68.8% in 1967 to 48.1% in 1983, and in the schoolchildren from 56.4% in 1967 to 8.2% in 1983. The reduction rate was higher in the riverside-delta area than in the inland area. 3) In the prevalence rate by age, 11.9% was first seen in the 5-9 age group and the rate gradually increased up to 75.0% in the 50-59 age group. By sex, the rate was higher in the male than in the female in the 20-29 age group and over. 4) In the natural transition of the prevalence rate by age, the reduction rate of the infection during the past 16 years was greater in the younger age groups up to the 40-49 age group and reached the same level in the age group 50-59. Reduction was seen again in the age group over 60s. By sex, the reduction rate was greater in the female than in the male in the 20-29 age group and over. By area, the reduction rate was greater in the riverside delta area than in the inland area, particularly in the young age groups. 5) In the intensity of the infection among the cases, the mean egg out-put per mg feces per infected cases(EPmg) in the inhabitants was 6.3. EPmg of those of the river-side-delta area was 15.4 and that of the in-land was 2.8. On the other hand, in the schoolchildren, EPmg was 3.2, and no difference was seen between the two areas, the river-side-delta area and the inland area. 6) In the transition of the intensity of the infection by area, EPmg among the inhabitants inexplically increased from 7.8 in 1967 to 15.4 in 1983. This was probably caused by uneven specimen collection in the process of sampling the population. EPmg of the inhabitants in the inland area and those of the schoolchildren of both riverside delta and inland areas showed a similar decrease in the past 16 years. 7) The intensity of the infection by age showed a relatively low level in the 20-29 age group and below, and EPmg 5.1-9.5 was seen in the 30-39 age group and over. Sex, Epmg was 5.8 in the male and 4.7 the female. By in 8) In the transition of the intensity of the infection, EPmg decreased from 6.2 in 1967 to 5.4 in 1983. By age, in contrast to the figures of 1967 in which EPmg gradually increased with some fluctuation from 1.1 in the 0-4 age group to peak 10.5 in the 50-59 age group, in 1983 lower intensity of the infection was seen in the age group from 10-14 to 20-29 with the EPmg range of 0.6-2.7. 9) In the distribution of the clonorchiasis cases by the range of EPmg value, 43.2% of the cases were in 0.1 0.9 and 34.6% in 1.0-4.9. As a whole by cumulative percent, 44.6% of them were under 0.9 as light infection and 86.1% of them under 9.9 up to moderate infection. By sex, no difference was seen in Epmg. 10) In the transition of the distribution by the range of Epmg, the cases were distributed up to the range 80.0-99.9 in 1967 and to 60.0-79.9 in 1983. By cumulative percent, in the range of 0.1-0.9 and less, light infection, 34.3% of them were distributed in 1967 and 44.6% in 1983 with about 10% increase. In the range of 5.0-9.9 and less, up to moderate infection, 83.2% in 1967 and 86.1% in 1983 of the cases were seen, respectively. 11) The practice of raw freshwater fish consumption among the inhabitants seems to have decreased in recent years. Those who admitted to raw freshwater fish consumption in the last two years among the infected inhabitants were 59.3%, although 86.8% of them professed to have experience with raw freshwater fish consumption. 31.7% of those who have had experience of the raw freshwater fish consumption denied any further consumption in recent years. From an interview of 543 school-children, 24.1% of them admitted to an experience of raw freshwater fish consumption. However, those who have practised in the past two years comprized 17.9%. Those who denied raw freshwater fish consumption in recent years among those who had such experience were 26.0% out of 131 interviewed. The rate of raw freshwater fish consumption in both inhabitants and schoolchildren were higher in the male than in the female. On the contrary, the rate of those who did not practise in recent years among those who had experience of raw freshwater fish consumption was higher in the female than in the male. 12) The major reason for the reduction of raw freshwater fish consumption among the local inhabitants was the risk of the fluke infection. However, it has become apparent that such change of taste has resulted from water pollution impact which has affected throughout the areas of the freshwater systems in this locality since last several years. 13) In animal survey, Clonorchis infection was seen in 14.8% of 88 dogs examined and 3.7% of 27 house rats examined. It was noted that populations of dogs and cats have increased in the villages surveyed. Although the prevalence rate was lower in the present survey than those of 1967, the significance of the animals as the reservoir host has not changed. 14) Prevalence rate of Clonorchis infection by cercariae in the first intermediate host, Parafossarulus manchouricus, was 0.6% out of 517 snails examined. The infection rate was lower in comparison with 2.3% out of 2,124 examined in 1967. Moreover, sharp decreases in number and distribution of the intermediate host snails in many watershed areas of the huge freshwater systems in this locality seemed to reduce transmission of Clonorchis in connection with the intermediate host stage of its life cycle. 15) Clonorchis infection in the second intermediate fish hosts was relatively low. The mean number of Clonorchis metacercaria per fish in Pseudorasbora parva was 517 in 1983, whereas it was 1943 in 1968 through 1969. Environmental water pollution has also caused the decreased fish population density in these areas, and this has also apparently affected to the practice of raw freshwater fish consumption among the local inhabitants. 16) In conclusion, endemicity of Clonorchis infection in Gimhae Gum and delta area of the Nagdong River has sharply decreased during the past 16 years. The major cause of the regressive transition of the infection was the water pollution of the land water systems of this locality. The pollution has upset the ecosystems comprizing of the intermediate hosts of Clonorchis in many areas, and also affected to a significant extent to the discontinuance of the local inhabitants for raw freshwater fish consumption.

• Parasites, Hosts and Diseases
• /
• v.29 no.3
• /
• pp.279-292
• /
• 1991

The immune response to sheep red blood cell (sRBC) was monitored in the mice infected with Ascaris strum or Trichinella spiralis. The effects of the infection with T. spiralis or the injection with cyclophosphamide (CY) as an immunosuppression agent prior to challenge infection with the embryonated eggs of A. suum were monitored in mice by means of the level of infection with A. strum and cellular and humoral immune response to sRBC. following the oral administration of 1, 000 eggs of A. suum to mice, delayed-type hypersensitivity (DTH) and rosette-forming rate were gradually decreased and reached to the lowest levels at the 5th week and 6th week postinfection, respectively, and then returned to normal at the loth week. The hemagglutinin (HA) and hemolysin (HE) titers were gradually elevated and reached to peak at the 3rd week postinfection, and then returned to normal level. The appearance ratios of the eosinophils and mast cells were in peak at the 4th week and the 2nd week postinfection, respectively. Meanwhile the harvest ratio of A. suum larvae from the liver and lungs was 21.97% at the 1st week postinfection. Following the oral administration of 300 T. spiralis infective larvae, DTH and rosette-forming rate were gradually decreased with the lapse of time and reached the lowest values in the 30th and 21st day of postinfection, and then slightly increased and transiently decreased in the 70th and 80th day of postinfection, respectively. HA and HE titers were the lowest in the 21st and 90th day, whereas the ratios of eosinophils and mast cells were the highest on the 40th and 14th day posti nfecti on, ruts petit i vela. Following the intraperitoneal injection of CY, the body weight, the spleen weight, DTH, rosette-orming ratio, HA and HE titers, the number of WBC and the ratio of the mast cell were predominantly decreased in the 5th day, and then returned to the same value of the 1st day postinjection. The ratio of eosinophils was gradually decreased following to advance of days. At the 1st, 5th and loth days after intraperitoneal injection of CY of 400 mg/kg, a dose with 1, 000 eggs of A. suum was administered orally to mice, and harvest rate of the larvae at the 7th day postadministration was 7.07% in the 1st day, 14.94% in the 5th day, 10.1% in the loth day, 8.02% in control group. The effect of prior infection with infective larvae of T. spiralis upon immunological sequelae of a challenge infection of mice with embryonated eggs of A. suum in 30 or 70 days interval was checked. On the 37th day of prior infection with T. spiralis, that was the 7th day with A. suum postinfection, DTH and rosette-forming rate were drastically decreased, but the ratio of mast cells was highly increased and the ratio of eosinophils, HA and HE titers were fairly increased. On the other hand, the rate of larvae harvest was 9.3% in experimental group in contrast with 22.18% in control group. Meanwhile the effect of immune response to sRBC was similar to that of the former, but DTH and rosettt-forming rate were greatly decreased in the 77th day after prior infection with the 7th day after challenge infection in compariton with control. At that time, Ascaris larvae were harvested 8.3% in experimental group in comparison with 10.5% in control group.

• Journal of Korean Home Economics Education Association
• /
• v.26 no.4
• /
• pp.115-128
• /
• 2014

The purpose of this study was to investigate home economics teachers' perception of cultural diversity education, to provide an efficient educational material for the multicultural education in teacher education and teacher retraining. 160 Home economics teachers answered the survey questionnaires. To analyze the data, SPSS 19.0 for Windows was used to conduct frequency analysis, factorial analysis, credibility analysis, t-test, one-way ANOVA, and Duncan's multiple comparison. The results of this study were as follows. Four dimensions of cultural diversity education were derived by factor analysis: cultural equality, diversity implementation, diversity value, comfort with diversity. As for their awareness about cultural diversity education, it was in the order of cultural equality, followed by diversity implementation, diversity value, and comfort with diversity. The groups were significantly different according to demographic variables. As for the whole awareness about cultural diversity education and the diversity implementation, group of age 40 teachers recognized more highly than other groups. Furthermore, teachers outside Jeonbuk area recognized more highly the cultural equality, diversity implementation, diversity value than those in Jeonbuk, which is the 3rd high area in the nation of multicultural family proportion. As for cultural equality and diversity implementation, teachers over 15 years of experience, recognized more highly than other groups. Those with the teacher certification in the college of education, recognized more highly the cultural equality, diversity value, comfort with diversity than teachers from the other colleges. Teachers who need multicultural education, recognized more highly cultural equality, diversity implementation, awareness of diversity than those who don't. These results imply that in home economics education, there must be more systematic studies on school field education and related educational programs in order to revitalize multicultural education. And for teachers with highly recognizing cultural diversity to conduct a systematic multicultural education more efficiently, there should be both systematic pre-service education programs at college level and in-service education programs for the teachers in terms of cultural diversity education.

• Journal of Nutrition and Health
• /
• v.45 no.5
• /
• pp.479-488
• /
• 2012

This study aims to investigate the nutritional status of night eaters using the data from 2005 Korean National Health and Nutrition Examination Survey. A total of 3,903 subjects aged 20 and above were divided into 3 groups by using 24-hr recall data according to the night snack calorie intake: non-night snack, night snack less than 500 kcal and 500 kcal and more. Their data were analyzed to find out the difference on the socio-demographic, anthropometric, blood pressure, blood parameters and dietary characteristics. Among the subjects, non-night eaters were 66.0%, night eaters of less than 500 kcal were 28.4% and 500 kcal and above were 5.6%. Male adults, young-aged, higher educated, higher income earner, breakfast skipper and frequent dine-outer (3 times and more a day) were found more in the night eaters with 500 kcal and above. Night eaters above 500 kcal showed higher waist circumference, Glu-FBS, Glu-PP120 and also showed higher daily intake of fat and alcohol per 1,000 kcal and food groups of meat & eggs, beverages and alcohol drinks (p < 0.05), while they showed the lowest carbohydrate energy ratio of $58.3{\pm}13.7$ among the three groups. These results suggest that habitual night snack intake above 500 kcal could lead to abdominal obesity and diabetes due to higher intake of meat, fat and alcohol.

• Journal of agricultural medicine and community health
• /
• v.12 no.1
• /
• pp.36-53
• /
• 1987

It is utmostly important to establish the efficient fitable way of peoples' active participation in primary health care especially in the areas where the public or governmental service input for the basic health care is insufficient like as in rural areas of Korea. In light of above reason, this study focused mainly on the evaluation of roles and activities of village health workers (VHWs) who were selected from grass- root level of village people in order to derive further motivation for active participation. This is believed to be a sort of feedback mechanisms. Actually, the authors collected the activity reports of VHWs who had been devoting themselves in the primary health care services of Jeomdong Area, of Yeoju Gun one of Korea University Community Health Action Programmes and survey record on the VHWs activity from correspondent people. 1 hose data were analyzed through computer programmed package. The activities performed by VHWs were limited to the performance in 1985 for conveniance. The summarized results were as follows; 1) General characteristics of VHWs. Among a total of 28 VHWs in the area, about 39.3g of them have been replaced up to the date since the implementation in 1983, because of moving out, occupational employment and of others. The age of majority (75.0%) lied between the range of 30-50, and educational background of 67.9% belonged to category of primary school graduation, about 50% of them experienced to be or were also entiled "chief of women club" of corresponding villages. 2) Work-load of VHWs. Each VHW was assigned for tasks of health care for average 55 households of 248 persons. They shared approximately 6 days a month for the activity in average and it covered 17 cases of basic health care in a month. A half of the VHWs performed home visits irregularly without solidified schedule. 3) Work performance analysis. Informations collected through VHWs were compared with data from official vital registration at local administration center "Myon Office" in 1985. VHWs collected 100.8 of new born, 116.2 of death, 58.3 of move in and 74.8 of move out in comparison with 100.0 of official registration each. Pregnant women of 79.8% of mothers among the total pregnancy of 94 which were confirmed as normally delivered or aborted cases by all means afterwards had been detected by VHWs as being pregnant and all of them received some of antenatal cares by VHWs. All(100%) of delivered women were detected by VHWs through home visits and they were cared postnatally. Whereas, according to the records of birth registration, the places of delivery were clinic in 33.7%, and mother's home in 66.3%, VHWs reported them to be clinic in 48.9%, midwifery in 20.2%. It was cleared that most of misinformation was caused by uncautious filling of birth registration at notification. Among the total of 717 eligible women under age 44 years, family planning status of 92.6% was reported by VHWs confirming practice of control to be 70.8% of reported fertile women. 4) Attitude of VHW on the roles and functions. Although 92.0% of VHWs expressed VHWs to be worthwhile, only 52.0% of them had dignity and satisfaction in their activity and 44.0% of them had passive attitude of working saying they followed direction regardlessly. Concerning difficulties in performance as a VHW, 60.7% of them pointed out lacking of medical and health related knowledge by themselves. Still, 64.0% of them thought visiting unfamilier house to be awful and 40.0% complained forms of activity to be difficult and hard. It was also revealed that 56.6% confessed lack of interest on community health service itself. Most of VHWs needed more educational training especially on clinical fields such as cares of gynecological diseases, hypertension, diabetes, and other chronic diseaes of the aged. Regular on-the-job basic trainings were said to be needed twice a year.

• Journal of Intelligence and Information Systems
• /
• v.27 no.2
• /
• pp.1-15
• /
• 2021

If it is difficult for us to make decisions, we ask for advice from friends or people around us. When we decide to buy products online, we read anonymous reviews and buy them. With the advent of the Data-driven era, IT technology's development is spilling out many data from individuals to objects. Companies or individuals have accumulated, processed, and analyzed such a large amount of data that they can now make decisions or execute directly using data that used to depend on experts. Nowadays, the recommender system plays a vital role in determining the user's preferences to purchase goods and uses a recommender system to induce clicks on web services (Facebook, Amazon, Netflix, Youtube). For example, Youtube's recommender system, which is used by 1 billion people worldwide every month, includes videos that users like, "like" and videos they watched. Recommended system research is deeply linked to practical business. Therefore, many researchers are interested in building better solutions. Recommender systems use the information obtained from their users to generate recommendations because the development of the provided recommender systems requires information on items that are likely to be preferred by the user. We began to trust patterns and rules derived from data rather than empirical intuition through the recommender systems. The capacity and development of data have led machine learning to develop deep learning. However, such recommender systems are not all solutions. Proceeding with the recommender systems, there should be no scarcity in all data and a sufficient amount. Also, it requires detailed information about the individual. The recommender systems work correctly when these conditions operate. The recommender systems become a complex problem for both consumers and sellers when the interaction log is insufficient. Because the seller's perspective needs to make recommendations at a personal level to the consumer and receive appropriate recommendations with reliable data from the consumer's perspective. In this paper, to improve the accuracy problem for "appropriate recommendation" to consumers, the recommender systems are proposed in combination with context-based deep learning. This research is to combine user-based data to create hybrid Recommender Systems. The hybrid approach developed is not a collaborative type of Recommender Systems, but a collaborative extension that integrates user data with deep learning. Customer review data were used for the data set. Consumers buy products in online shopping malls and then evaluate product reviews. Rating reviews are based on reviews from buyers who have already purchased, giving users confidence before purchasing the product. However, the recommendation system mainly uses scores or ratings rather than reviews to suggest items purchased by many users. In fact, consumer reviews include product opinions and user sentiment that will be spent on evaluation. By incorporating these parts into the study, this paper aims to improve the recommendation system. This study is an algorithm used when individuals have difficulty in selecting an item. Consumer reviews and record patterns made it possible to rely on recommendations appropriately. The algorithm implements a recommendation system through collaborative filtering. This study's predictive accuracy is measured by Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). Netflix is strategically using the referral system in its programs through competitions that reduce RMSE every year, making fair use of predictive accuracy. Research on hybrid recommender systems combining the NLP approach for personalization recommender systems, deep learning base, etc. has been increasing. Among NLP studies, sentiment analysis began to take shape in the mid-2000s as user review data increased. Sentiment analysis is a text classification task based on machine learning. The machine learning-based sentiment analysis has a disadvantage in that it is difficult to identify the review's information expression because it is challenging to consider the text's characteristics. In this study, we propose a deep learning recommender system that utilizes BERT's sentiment analysis by minimizing the disadvantages of machine learning. This study offers a deep learning recommender system that uses BERT's sentiment analysis by reducing the disadvantages of machine learning. The comparison model was performed through a recommender system based on Naive-CF(collaborative filtering), SVD(singular value decomposition)-CF, MF(matrix factorization)-CF, BPR-MF(Bayesian personalized ranking matrix factorization)-CF, LSTM, CNN-LSTM, GRU(Gated Recurrent Units). As a result of the experiment, the recommender system based on BERT was the best.

• Journal of the Korean earth science society
• /
• v.23 no.3
• /
• pp.280-293
• /
• 2002

In this study, we analyzed the long-term distribution patterns of $CH_4$ determined from the Moo-Ahn (MAN) observatory in relation with those derived from the world major background monitoring sites. Comparison of the data were made using those data sets collected for the period between Aug. 1995 to Dec. 1991. The mean $CH_4$ concentration of MAN observatory was measured to be 1898${\pm}$85.3 ppb, recording the highest concentration of all the monitoring sites. When the concentration of $CH_4$ for different stations was compared over latitudinal scale, its concentration appeared to increase systematically as a function of latitude with an exception of MAN (and the other Korean monitoring site at Tae Ahn). Moreover, such phenomenon was more distinctive in Northern than Southern Hemisphere. According to the analysis of the monthly distribution patterns of $CH_4$ at MAN observatory, its concentration level began to increase from the months of February/March and peaked during August. In addition, when the level of oscillation in monthly concentrations (between the maximum and minimum values) was checked, differences were significant between MAN and other monitoring stations. If the rate of concentration change was checked using the data sets collected for this limited time period in terms of linear regression analysis, results for MAN showed the highest annual increasing rate of 16.5 ppb. It is hence suggested that the largest variability in the $CH_4$ distribution patterns at MAN observatory may be reflected by the high irregularity in its source/sink processes.

• 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.