• Journal of Intelligence and Information Systems
• /
• v.19 no.2
• /
• pp.125-140
• /
• 2013

We call a data set in which the number of records belonging to a certain class far outnumbers the number of records belonging to the other class, 'imbalanced data set'. Most of the classification techniques perform poorly on imbalanced data sets. When we evaluate the performance of a certain classification technique, we need to measure not only 'accuracy' but also 'sensitivity' and 'specificity'. In a customer churn prediction problem, 'retention' records account for the majority class, and 'churn' records account for the minority class. Sensitivity measures the proportion of actual retentions which are correctly identified as such. Specificity measures the proportion of churns which are correctly identified as such. The poor performance of the classification techniques on imbalanced data sets is due to the low value of specificity. Many previous researches on imbalanced data sets employed 'oversampling' technique where members of the minority class are sampled more than those of the majority class in order to make a relatively balanced data set. When a classification model is constructed using this oversampled balanced data set, specificity can be improved but sensitivity will be decreased. In this research, we developed a hybrid model of support vector machine (SVM), artificial neural network (ANN) and decision tree, that improves specificity while maintaining sensitivity. We named this hybrid model 'hybrid SVM model.' The process of construction and prediction of our hybrid SVM model is as follows. By oversampling from the original imbalanced data set, a balanced data set is prepared. SVM_I model and ANN_I model are constructed using the imbalanced data set, and SVM_B model is constructed using the balanced data set. SVM_I model is superior in sensitivity and SVM_B model is superior in specificity. For a record on which both SVM_I model and SVM_B model make the same prediction, that prediction becomes the final solution. If they make different prediction, the final solution is determined by the discrimination rules obtained by ANN and decision tree. For a record on which SVM_I model and SVM_B model make different predictions, a decision tree model is constructed using ANN_I output value as input and actual retention or churn as target. We obtained the following two discrimination rules: 'IF ANN_I output value <0.285, THEN Final Solution = Retention' and 'IF ANN_I output value ${\geq}0.285$, THEN Final Solution = Churn.' The threshold 0.285 is the value optimized for the data used in this research. The result we present in this research is the structure or framework of our hybrid SVM model, not a specific threshold value such as 0.285. Therefore, the threshold value in the above discrimination rules can be changed to any value depending on the data. In order to evaluate the performance of our hybrid SVM model, we used the 'churn data set' in UCI Machine Learning Repository, that consists of 85% retention customers and 15% churn customers. Accuracy of the hybrid SVM model is 91.08% that is better than that of SVM_I model or SVM_B model. The points worth noticing here are its sensitivity, 95.02%, and specificity, 69.24%. The sensitivity of SVM_I model is 94.65%, and the specificity of SVM_B model is 67.00%. Therefore the hybrid SVM model developed in this research improves the specificity of SVM_B model while maintaining the sensitivity of SVM_I model.

• Journal of Preventive Medicine and Public Health
• /
• v.26 no.4 s.44
• /
• pp.587-598
• /
• 1993

Hydrofluoric acid is one of the strongest irritating, corrosive and poisonous inorganic chemicals. Hydrofluoric acid burns are occurring with ever-increasing frequency due to the wide use of this acid in industries. Hydrofluoric acid burns are characterized by severe progressive tissue destruction and excruciating pain due to the unique properties of the freely dissolvable fluoride ion. The authors reviewed medical records of 32 cases (36 spells) of hydrofluoric acid burns which occurred in a hydrofluoric acid manufacturing factory from Sep. 1, 1990 to June 30, 1993. The results are as follows; 1. Eleven measurements of air concentrations of hydrofluoric acid by detection tube method from 1990 to 1992 were all below TLV (Department of Labor, R.O. K). 2. There were 19 cases (22 spells) of hydrofluoric acid burns which occurred during the study period among regular employees. The overall incidence density of hydrofluoric acid was 17.8 cases (20.6 spells) per 100 person-year. Incidence density was 19.0 cases (22.0 spells) per 100 person-year among male workers and there were no female cases. Incidence density was 32.9 cases (38.3 spells) per 100 person-year among production workers and 1.9 cases (1.9 spells) per 100 person-years among management workers with the difference being statistically significant (P<0.01). 3. Of 32 cases (36 spells) of hydrofluoric acid burns among workers who were regularly employed or temporarily employed, 26 spells (81.2%) were between age 20 to 39. In 15 spells(41.7%) burns occurred between 12:00 and 17:59 with 16 spells(44.3%) having arrived at hospital within 2 hours after the accident. 4. Of 36 spells, the main cause of hydrofluoric acid burns were by splashes (8 spells, 22.2 %). The most frequent site of burns were fingers and pain was the most frequent symptom. Thirty spells (83.3%) of the hydrofluoric acid burns were treated with local injection of antidote (calcium gluconate). Complete recovery without scarring were observed in most of the cases (34 out of 36 cases, 94.4%). The study results suggest that to prevent hydrofluoric acid burns, environmental control and the wearing of hydrofluoric acid resistant protective clothes and gloves are important. It is also stressed that establishment of an emergency management and a transfer system for hydrofluoric acid burn victims is necessary.

• Journal of Navigation and Port Research
• /
• v.31 no.8
• /
• pp.629-635
• /
• 2007

Reduced crew performance is frequently cited as a major causal factor in maritime accident causation. Although considerable research has been conducted on the hours of work undertaken by seafarers through interviews and the analysis of records, experimental studies to observe the effects of factors such as high workload, shift patterns, stress, sleep deprivation and disturbance on the cognitive performance of mariners have been limited. Other safety-critical transport industries, such as aviation and rail, have developed fatigue management tools to help manage the work patterns of their operators. Such a tool for mariners would assist shipboard crew, marine pilots and shore management in planning and improving work schedules. The overall aim of this paper is to determine a fatigue factor, which can be applied to human performance data, as part of a software program that calculates total cognitive performance. This program enables us to establish the levels of cognitive performance of a group of marine pilots to test a decision-making task based on radar information. This paper addresses one of the factors that may contribute to the determination of various fatigue factors: the effect of different work patterns on the cognitive performance of a marine pilot.

• Journal of Animal Science and Technology
• /
• v.48 no.6
• /
• pp.767-776
• /
• 2006

Corresponding author:Deukhwan Lee, Department of Animal Life Resources, Hankyong National Univ. Seokjeong-dong Anseong-si Gyeonggi-do, Rep. of Korea 456-749Tel: 031-670-5091, Fax: 031-676-5091, E-mail : dhlee@hknu.ac.krGeneral performance with including environmental and management effects on linear type traits in Holstein cows were investigated. 115,646 valid records measured from cows over 1 yrs of age by Korean Animal Improvement Association from 2000 to 2004 were used for this study. Farm, appraisal year-month, appraisal person should affect linear type and scoring traits. Most of type traits and scoring traits would be significantly affected by parity and lactation stage after absorbing farm-appraisal year-month-person effects. Otherwise, some traits such as traits related to udder would be affected by registration criteria. However, interval of appraisal time and milking time would not affect these traits. The scores related to udder, teat placement and foot angle would be positively related to parity. Final score would be optimized at cows of 2nd and 3rd parities. Dairy form, front teat placement, rear udder height, rear udder width and final score would show similar to pattern of lactation curve at lactation stage. Dairy capacity composite index would also show similar to lactation curve. This result would be indicated that more concise standardizing system for linear type scoring rules would be needed. Furthermore, correcting system for parity and lactation stage would be needed because this factor should affect physiological status, specially, udder status.

• Journal of Preventive Medicine and Public Health
• /
• v.35 no.4
• /
• pp.340-350
• /
• 2002

Objectives : To evaluate the annual visit days, the annual prescription days and the medical costs of hypertensive patients. Methods : The medical insurance records of 40,267 incident patients with the diagnostic code of hypertension from September 1998 through August 1999 in Daegu city were reviewed. Results : The proportion of the most proper medical care pattern group (Group VIII) who visited for 6-15 days with 240 prescription days or more a year was only 6.2%. The proper care group (Group IX) who visited for more than 16 days with 240 prescription days or more a year was 9.3%. The overall proper care group (Group VIII+IX) was therefore 15.5%. The proportion of the insufficient care group (Group I, IV) in both the number of visiting days and prescription days was 57.4%. The mean prescription day of the most proper group (Group VIII) was 29 days; the mean annual medical expenses,453,587won; the mean annual amount paid by patients, 218,013won; and mean medical expenses per prescription day, 1,483won. The proportion of the overall proper care group (Group VIII+IX) was significantly higher in adults aged 50-59, those who were enrolled in industrial workers health insurance as well as government employees and private school teachers health insurance, and those who made a higher contribution per month (p<0.01). According to the type of medical facilities, the proportion of the most proper medical care pattern group was highest in the general hospitals (9.3%) but the overall proper care group was higher in the public health centers (22.1%) and private clinics (17.1%). Conclusions : The management system of hypertension should be reinforced urgently. Therefore, it is necessary to develop guidelines including the number of visiting days per year and prescription days per visit day, and make the system provide medical facilities to more properly care for hypertensive patients.

• The Korean Journal of Archival Studies
• /
• no.46
• /
• pp.153-187
• /
• 2015

This study aims to discuss issues about non-existent information of the information disclosure system and to provide alternative strategies for the issues. For the theoretical discussion it reviews the definitions and standards of non-existent information and analyzes legal aspects and statistical changes of non-existent information. Furthemore, in order to discuss a current status and problems of non-existent information at the central administrative agencies, it analyzes the cases of the non-existent information notification. According to analysis results, non-existent information status of the surveyed institutions is a total of 4,421 cases for three years and it shows the continuous increasing trend year after year. The number of institutions that have the number of non-existent information equal to the number of nondisclosures or over it reached about 40%. It means excluding non-existent information from the reasons of nondisclosure influenced disclose rates and nondisclosure rates of many agencies. In the type analysis of the non-existent information reasons, the most main reason, the case of not producing or receiving the requested information by public institutions takes over 75% among the whole reasons. The next reason is the case of collecting or processing information takes over 7-10%. This study found the operational issues, as analyzing notifications of non-existent information. The operational issues are 1) the incomplete explanation of non-existent information, 2) the unclear scope of the collection and processing, 3) the problem of the transfer processing, and 4) the problem of recording management. Therefore, this study suggested some improvements of the perspective and the technical and procedural aspects. First, information disclosure issues including non-existent information are to be understood as an extension of records management. Second, disclosure service should improve overall based on advanced understanding. Third, the management procedures of non-existent information should be improved. Fourth, specific guidelines for handling non-existent information should be developed.

• Journal of Intelligence and Information Systems
• /
• v.17 no.2
• /
• pp.77-96
• /
• 2011

An exhibition is defined as market events for specific duration to present exhibitors' main products to business or private visitors, and it plays a key role as effective marketing channels. As the importance of exhibition is getting more and more, domestic exhibition industry has achieved such a great quantitative growth. But, In contrast to the quantitative growth of domestic exhibition industry, the qualitative growth of Exhibition has not achieved competent growth. In order to improve the quality of exhibition, we need to understand the preference or behavior characteristics of visitors and to increase the level of visitors' attention and satisfaction through the understanding of visitors. So, in this paper, we used the observation survey method which is a kind of field research to understand visitors and collect the real data for the analysis of behavior pattern. And this research proposed the following methodology framework consisting of three steps. First step is to select a suitable exhibition to apply for our method. Second step is to implement the observation survey method. And we collect the real data for further analysis. In this paper, we conducted the observation survey method to obtain the real data of the KIDS & EDU EXPO for Children in SETEC. Our methodology was conducted on 160 visitors and 78 booths from November 4th to 6th in 2010. And, the last step is to analyze the record data through observation. In this step, we analyze the feature of exhibition using Demographic Characteristics collected by observation survey method at first. And then we analyze the individual booth features by the records of visited booth. Through the analysis of individual booth features, we can figure out what kind of events attract the attention of visitors and what kind of marketing activities affect the behavior pattern of visitors. But, since previous research considered only individual features influenced by exhibition, the research about the correlation among features is not performed much. So, in this research, additional analysis is carried out to supplement the existing research with data mining techniques. And we analyze the relation among booths using data mining techniques to know behavior patterns of visitors. Among data mining techniques, we make use of two data mining techniques, such as clustering analysis and ARM(Association Rule Mining) analysis. In clustering analysis, we use K-means algorithm to figure out the correlation among booths. Through data mining techniques, we figure out that there are two important features to affect visitors' behavior patterns in exhibition. One is the geographical features of booths. The other is the exhibit contents of booths. Those features are considered when the organizer of exhibition plans next exhibition. Therefore, the results of our analysis are expected to provide guideline to understanding visitors and some valuable insights for the exhibition from the earlier phases of exhibition planning. Also, this research would be a good way to increase the quality of visitor satisfaction. Visitors' movement paths, booth location, and distances between each booth are considered to plan next exhibition in advance. This research was conducted at the KIDS & EDU EXPO for Children in SETEC(Seoul Trade Exhibition & Convention), but it has some constraints to be applied directly to other exhibitions. Also, the results were derived from a limited number of data samples. In order to obtain more accurate and reliable results, it is necessary to conduct more experiments based on larger data samples and exhibitions on a variety of genres.

• Korean journal of food and cookery science
• /
• v.26 no.5
• /
• pp.603-616
• /
• 2010

Based on recent dramatic increases in foodborne outbreaks in restaurants, self-managed sanitation systems are now recommended to control contributing risk factors. This study aimed to improve sanitation management practices in restaurants and had two objectives. First, we tried to develop a self-managed sanitation check-list, including risk factors contributing to foodborne illness and Korean food hygiene regulation articles. We also tried to evaluate current sanitation management practices in restaurants according to operation and restaurant type. Thirty restaurants were evaluated by on-the-spot inspectors using an auditing tool consisting of four dimensions, seventeen categories, and forty-one items. Total compliance rate categorized by operation type significantly differed between chain restaurants and self-managed restaurants, with values of 85.5% and 51.6%, respectively. Therefore, self-managed restaurants, which showed the lowest compliance rate of below 30.0%, need more strict control to improve current unsanitary management practices, specifically relating to 'sterilization of knives, chopping boards, and wiping cloths', 'sanitation training', 'not allowing access into the kitchen to outsiders', 'handling of food or utensils on shelves at a 15 cm distance away from floor', 'prevention of cross-contamination of cooked foods or vegetables', and 'records of kitchen access or inspection'. Thus, an effective food sanitation system is essential and should be implemented to improve the existing sanitary conditions in restaurants. However, the most important factor to achieving food sanitation management objectives is food handlers' self-motivation.

• Journal of Korea Technology Innovation Society
• /
• v.13 no.3
• /
• pp.538-557
• /
• 2010

Korean Government has supplied small-medium enterprises with various technology financing programs such as R&D grant." loans, equities, guarantees since 1980's. The purpose of this study is to render policy suggestions for policy makers by analysing problems with regard to regulations. The suggestions of this study are such as the following. First, the current three months of the fund formation period should be extended within six months for general partners to have enough time to raise funds in the current tough capital market situation. Second, the government's non priority loss devotion system should keep going to prevent general partner's moral hazard. Third, in order to fulfill the policy purpose that is to create globally well-known companies, the government investment ratio should scale up from the current 20% to 30% of the total fund size in that the stable capital supply for core technology commercialization, industrialization and globalization of companies is achieved. Finally, general partners of the New Growth Engine Fund should showoff their investment abilities to the potential investors such as institutional investors by accumulating all track records and know-hows that prove their specialities and implementing action plans to achieve the best result of their investment. This study provides general considerations of the selection procedure, the government investment ratio, the government priority loss devotion system and so on when formating the public funds.

• Journal of the Society of Disaster Information
• /
• v.11 no.4
• /
• pp.622-633
• /
• 2015

With recent news reports on violent crimes committed by foreigners, public hatred toward foreign criminals and fear about such crimes are growing, calling into question public safety measures implemented by the police and the immigration system. At the same time, due to globalization and structural changes in the labor market, it is certain that the influx of migrant workers will continue to increase. Against this backdrop, this paper seeks to examine the current status of foreign nationals staying in Korea and their criminal records focusing on Suwon city, and propose public safety measures the police could take with regard to crimes committed by foreigners. Also, the paper looks into public safety measures of the Korea Immigration Service and Suwon city, and suggests short- and long-term measures to prevent crimes committed by foreigners. The paper begins by examining the current public safety measures of the police and Suwon City concerning crimes committed by foreign nationals residing in Korea. Based on such examination, it points out insufficient patrol and investigation personnel as problems, and recommends short- and long-term measures for future improvement. This paper maintains that for the Korean society to successfully deal with the increasing number of foreigners in Korea, the current immigration system and public safety measures of the police will have to change. While the police alone cannot reduce the crime rates at local and national levels, nonetheless, proactive measures by the police are needed given the growing number of migrant workers and immigrants in Korea. Also important are mindset changes not only of relevant authorities, but also of the public. The paper recommends a set of short-and long-term public safety measures the police should institute to efficiently deal with crimes committed by foreigners.