• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.175-197
• /
• 2021

Recently, with the development of deep learning technology, research on unstructured data analysis is being actively conducted, and it is showing remarkable results in various fields such as classification, summary, and generation. Among various text analysis fields, text classification is the most widely used technology in academia and industry. Text classification includes binary class classification with one label among two classes, multi-class classification with one label among several classes, and multi-label classification with multiple labels among several classes. In particular, multi-label classification requires a different training method from binary class classification and multi-class classification because of the characteristic of having multiple labels. In addition, since the number of labels to be predicted increases as the number of labels and classes increases, there is a limitation in that performance improvement is difficult due to an increase in prediction difficulty. To overcome these limitations, (i) compressing the initially given high-dimensional label space into a low-dimensional latent label space, (ii) after performing training to predict the compressed label, (iii) restoring the predicted label to the high-dimensional original label space, research on label embedding is being actively conducted. Typical label embedding techniques include Principal Label Space Transformation (PLST), Multi-Label Classification via Boolean Matrix Decomposition (MLC-BMaD), and Bayesian Multi-Label Compressed Sensing (BML-CS). However, since these techniques consider only the linear relationship between labels or compress the labels by random transformation, it is difficult to understand the non-linear relationship between labels, so there is a limitation in that it is not possible to create a latent label space sufficiently containing the information of the original label. Recently, there have been increasing attempts to improve performance by applying deep learning technology to label embedding. Label embedding using an autoencoder, a deep learning model that is effective for data compression and restoration, is representative. However, the traditional autoencoder-based label embedding has a limitation in that a large amount of information loss occurs when compressing a high-dimensional label space having a myriad of classes into a low-dimensional latent label space. This can be found in the gradient loss problem that occurs in the backpropagation process of learning. To solve this problem, skip connection was devised, and by adding the input of the layer to the output to prevent gradient loss during backpropagation, efficient learning is possible even when the layer is deep. Skip connection is mainly used for image feature extraction in convolutional neural networks, but studies using skip connection in autoencoder or label embedding process are still lacking. Therefore, in this study, we propose an autoencoder-based label embedding methodology in which skip connections are added to each of the encoder and decoder to form a low-dimensional latent label space that reflects the information of the high-dimensional label space well. In addition, the proposed methodology was applied to actual paper keywords to derive the high-dimensional keyword label space and the low-dimensional latent label space. Using this, we conducted an experiment to predict the compressed keyword vector existing in the latent label space from the paper abstract and to evaluate the multi-label classification by restoring the predicted keyword vector back to the original label space. As a result, the accuracy, precision, recall, and F1 score used as performance indicators showed far superior performance in multi-label classification based on the proposed methodology compared to traditional multi-label classification methods. This can be seen that the low-dimensional latent label space derived through the proposed methodology well reflected the information of the high-dimensional label space, which ultimately led to the improvement of the performance of the multi-label classification itself. In addition, the utility of the proposed methodology was identified by comparing the performance of the proposed methodology according to the domain characteristics and the number of dimensions of the latent label space.

• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.157-173
• /
• 2021

With the development of information technology, the amount of available information increases daily. However, having access to so much information makes it difficult for users to easily find the information they seek. Users want a visualized system that reduces information retrieval and learning time, saving them from personally reading and judging all available information. As a result, recommendation systems are an increasingly important technologies that are essential to the business. Collaborative filtering is used in various fields with excellent performance because recommendations are made based on similar user interests and preferences. However, limitations do exist. Sparsity occurs when user-item preference information is insufficient, and is the main limitation of collaborative filtering. The evaluation value of the user item matrix may be distorted by the data depending on the popularity of the product, or there may be new users who have not yet evaluated the value. The lack of historical data to identify consumer preferences is referred to as data sparsity, and various methods have been studied to address these problems. However, most attempts to solve the sparsity problem are not optimal because they can only be applied when additional data such as users' personal information, social networks, or characteristics of items are included. Another problem is that real-world score data are mostly biased to high scores, resulting in severe imbalances. One cause of this imbalance distribution is the purchasing bias, in which only users with high product ratings purchase products, so those with low ratings are less likely to purchase products and thus do not leave negative product reviews. Due to these characteristics, unlike most users' actual preferences, reviews by users who purchase products are more likely to be positive. Therefore, the actual rating data is over-learned in many classes with high incidence due to its biased characteristics, distorting the market. Applying collaborative filtering to these imbalanced data leads to poor recommendation performance due to excessive learning of biased classes. Traditional oversampling techniques to address this problem are likely to cause overfitting because they repeat the same data, which acts as noise in learning, reducing recommendation performance. In addition, pre-processing methods for most existing data imbalance problems are designed and used for binary classes. Binary class imbalance techniques are difficult to apply to multi-class problems because they cannot model multi-class problems, such as objects at cross-class boundaries or objects overlapping multiple classes. To solve this problem, research has been conducted to convert and apply multi-class problems to binary class problems. However, simplification of multi-class problems can cause potential classification errors when combined with the results of classifiers learned from other sub-problems, resulting in loss of important information about relationships beyond the selected items. Therefore, it is necessary to develop more effective methods to address multi-class imbalance problems. We propose a collaborative filtering model using CGAN to generate realistic virtual data to populate the empty user-item matrix. Conditional vector y identify distributions for minority classes and generate data reflecting their characteristics. Collaborative filtering then maximizes the performance of the recommendation system via hyperparameter tuning. This process should improve the accuracy of the model by addressing the sparsity problem of collaborative filtering implementations while mitigating data imbalances arising from real data. Our model has superior recommendation performance over existing oversampling techniques and existing real-world data with data sparsity. SMOTE, Borderline SMOTE, SVM-SMOTE, ADASYN, and GAN were used as comparative models and we demonstrate the highest prediction accuracy on the RMSE and MAE evaluation scales. Through this study, oversampling based on deep learning will be able to further refine the performance of recommendation systems using actual data and be used to build business recommendation systems.

• Journal of Nutrition and Health
• /
• v.54 no.2
• /
• pp.139-151
• /
• 2021

Purpose: This study examines the correlation between fish consumption and the risk of mild cognitive impairment in the elderly living in rural areas. Methods: The Yangpyeong cohort data collected from Yangpyeong in July 2009 and August 2010 was used as the data set. Adults greater than or equal to 60 years who have completed the Korean version of the Mini-Mental State Examination (MMSE-KC) were selected for the study. After excluding participants with less than 500 kcal of energy intake (n = 2), a total of 806 adults were enrolled as the final subjects. Cognitive function was assessed using the MMSE-KC, and dietary intake was collected using the quantitative food frequency questionnaire comprising 106 foods or food groups. Results: The educational level, proportion of people who exercise, fruits and vegetable intake, and energy intake, tended to increase with fish intake among men, while increasing age resulted in decreased fish consumption. Among women, the educational level, proportion of subjects who exercise, proportion of subjects currently taking dietary supplements, fruits and vegetable intake, and energy intake, tended to increase with fish consumption, whereas increasing age showed decreasing fish consumption. Increased fish intake resulted in a higher MMSE-KC score after adjusting for the confounding variables in women (p for trend = 0.016), but no significant trend was observed between fish intake and MMSE-KC score in men. Fish intake was inversely related to the risk of mild cognitive impairment after adjusting for covariates in women (Q1 vs. Q4; odds ratio, 0.46 [0.23-0.90]; p for trend = 0.009). Conclusion: This study determined that increased fish consumption is correlated with reduced risk of mild cognitive impairment in the female elderly. Further longitudinal studies with larger samples are required to determine a causal relationship between fish intake and cognitive function.

• Korean Journal of Psychosomatic Medicine
• /
• v.28 no.2
• /
• pp.185-193
• /
• 2020

Objectives : The aim of this study was to develop and to verify the Korean version of the Eating Disorder Diagnosis Scale DSM-5 (K-EDDS) as a web-based diagnostic system, which enables rapid diagnosis of patients for early intervention. Methods : A total of 119 persons participated in the study, including patients with eating disorders (n=38) and college students (n=81). Along with the paper-and-pencil SCOFF, all participants completed the web-based K-EDDS, the Eating Disorder Examination-Questionaire (EDE-Q), and the Clinical Impairment Assessment Questionnaire (CIA). The semi-structured interview using the Eating Disorder Examination Interview (EDE) was conducted for participants with two or more SCOFF scores. Within two weeks, the web-based K-EDDS, the EDE-Q, and the CIA were re-tested. Results : In the exploratory factor analysis, four factors were extracted : body dissatisfaction, binge behaviors, binge frequency and compensatory behaviors. The four subscales of the web-based K-EDDS had significant correlation with each of the four subscales of the EDE-Q. The internal consistency of the web-based K-EDDS was highly satisfactory (Cronbach's alpha=0.93). The diagnostic agreement between the web-based K-EDDS and the EDE was excellent (96.83%), and the web-based K-EDDS's test-retest diagnostic agreement was fairly good (92.86%). The web-based K-EDDS and the CIA also showed significant differences between patients and general population, supporting discriminant validity. Conclusions : This study suggested that the web-based K-EDDS is a valid tool for assisting diagnosis of eating disorders based on DSM-5 in clinical and research fields.

• Journal of Nutrition and Health
• /
• v.54 no.1
• /
• pp.10-22
• /
• 2021

Purpose: Key Foods refers to foods that have a high contribution in the nutrient intake of individuals, and exert important effects on their health. This study was undertaken to identify Korean Key Foods, using data from the 7th Korea National Health and Nutrition Examination Survey (KNHNES). Methods: The data source for the extraction of Key Foods was the 24-hour dietary survey data obtained from the 7th KNHNES (2016-2018), and 21,271 subjects were evaluated. A total of 17 nutrients were selected as the key nutrients for identifying the Key Foods, including energy, carbohydrates, protein, lipid, dietary fiber, calcium, phosphorus, iron, sodium, potassium, vitamin A, thiamin, riboflavin, niacin, vitamin C, cholesterol, and sugars. The nutrient consumption approach was applied to generate a list of potential Key Foods. Foods included in 85% of the cumulative intake contribution from one or more key nutrients, were subsequently selected as Key Foods. Results: Of the 1,728 foods consumed by survey respondents, we extracted 728 Key Foods. These Key Foods explained 94% key nutrient intakes of the subjects. Based on the contribution rate to key nutrient intake, the top 10 Key Foods identified were multigrain rice (5.32%), plain white rice (4.23%), milk (3.3%), cabbage kimchi (2.82%), grilled pork belly (1.56%), apples (1.52%), fried eggs (1.49%), cereal (1.36%), instant coffee mix (1.21%), and sweet potatoes (1.12%). These 10 foods accounted for 23.93% total key nutrient intake of the survey respondents. Conclusion: Seven hundred and twenty-eight foods were extracted and identified as the 2020 Korean Key Foods. These Key Foods can be considered the priority foods to be analyzed for establishing a national nutrient database.

• Journal of Nutrition and Health
• /
• v.54 no.3
• /
• pp.292-304
• /
• 2021

Purpose: This study investigated the association between the mothers' use of home meal replacement (HMR) in their children's meals and the diet quality of their young children. Methods: Three hundred and thirty-seven mothers with five-year-old kindergartners in Sejong city participated in the survey from June to July 2020. The questionnaire consisted of the status of HMR use in children's meals and questions for assessing the nutrition quotient for preschoolers (NQ-P). The subjects were classified into three groups according to the frequency of HMR use in children's meals: using HMR more than three times a week (high-frequency group; [HG], n = 65), one-two times a week (moderate-frequency group; [MG], n = 145), and less than once a week (low-frequency group; [LG], n = 130). Results: The mothers' mean age was 38.3 years. The average monthly cost of purchasing HMRs was highest at 200,000-300,000 won in HG, 50,000-100,000 won in MG, and less than 50,000 won in LG (p < 0.001). The consumption frequency of processed meats, fast foods, processed beverages, and sweet & fatty snacks was significantly higher in the HG group than the other groups. The mean NQ-P score was 60.5 in HG, 63.0 in MG, and 64.5 in LG, showing a significant difference (p < 0.01). In the sub-score according to the three areas, there were no significant differences in balance and environment among the three groups. In the moderation area, however, the score was 44.1 in HG, 51.3 in MG, and 57.5 in LG Group, showing a significant difference (p < 0.001). Conclusion: The increase in HMR use was related to the decreased diet quality in the overall and moderation areas of children's diet. These results support the importance of nutrition education for mothers, which aims to reduce their children's access and exposure to processed foods, such as HMR.

• 한국노년학
• /
• v.37 no.1
• /
• pp.173-200
• /
• 2017

This study is designed to verify the actual effect on the prevention of the elderly suicide of the integrated crisis intervention service which has been widely provided across all local communities in Gyeonggi-province focusing on the integrated crisis intervention model developed for the prevention of elderly suicide. The integrated crisis intervention model for the local communities and its manual were developed for the prevention of elderly suicide by integrating the crisis intervention theory which contains local community's integrated system approach and the stress vulnerability theory. For the analysis of the effect, the geriatric depression and suicidal ideation scale was adopted and the data was collected as follows; The data was collected from 258 people in the first preliminary test. Then, it was collected from the secondary test of 184 people after the integrated crisis intervention service was performed for 6 months. The third collection of data was made from 124 people after 2 or 3 years later using the backward tracing method. As for the analysis, the researcher used the R Statistics computing to conduct the test equating, and the vertical scaling between measuring points. Then, the researcher conducted descriptive statistics analysis and univariate analysis of variance, and performed multi-level modeling analysis using Bayesian estimation. As a result of the study, it was found out that the integrated crisis intervention model which has been developed for the elderly suicide prevention has a statistically significant effect on the reduction of elderly suicide in terms of elderly depression and suicide ideation in the follow-up measurement after the implementation of crisis intervention rather than in the first preliminary scores. The integrated crisis intervention model for the prevention of elderly suicide was found to be effective to the extent of 0.56 for the reduction of depression and 0.39 for the reduction of suicidal ideation. However, it was found out in the backward tracing test conducted 2-3 years after the first crisis intervention that the improved values returned to its original state, thus showing that the effect of the intervention is not maintained for long. Multilevel analysis was conducted to find out the factors such as the service type(professional counseling, medication, peer counseling), characteristics of the client (sex, age), the characteristics of the counselor(age, career, major) and the interaction between the characteristics of the counselor and intervention which affect depression and suicidal ideation. It was found that only medication can significantly reduce suicidal ideation and that if the counselor's major is counseling, it significantly further reduces suicidal ideation by interacting with professional counseling. Furthermore, as the characteristics of the suicide prevention experts are found to regulate the intervention effect on elderly suicide prevention in applying integrated crisis intervention model, the primary consideration should be given to the counseling ability of these experts.

• Korean Journal of Psychosomatic Medicine
• /
• v.29 no.1
• /
• pp.49-57
• /
• 2021

Objectives : This study evaluated the medical communication skills of trainee doctors and analyzed the relationship between medical communication skills, self-efficacy on clinical performance (SECP) and empathy. Methods : A total of 106 trainee doctors from a university hospital participated. The questionnaire comprised self-evaluated medical communication skills, modified SECP and the Korean version of the Jefferson Scale of Empathy-Health Professionals version. The mean difference in medical communication skills scores according to gender, age, division (intern, internal medicine group or surgery group) and position (intern, first-/second- and third-/fourth-year residents) were analyzed. Pearson correlation coefficients were determined between medical communication skills, modified SECP and empathy. The effects of each variable on medical communication skills were verified using the structural equation model. Results : There were no statistically significant mean differences in self-evaluated medical communication skills according to gender, age, division or position. Medical communication skills had a significant positive correlation with modified SECP (r=0.782, p<0.001) and empathy (r=0.210, p=0.038). Empathy had a direct effect on modified SECP (β=0.30, p<0.01) and modified SECP had a direct effect on medical communication skills (β=0.80, p<0.001). Empathy indirectly influenced medical communication skills, mediating modified SECP (β=0.26, p<0.05). Conclusions : Medical communication skills are an important core curriculum of residency programs, as they have a direct correlation with SECP, which is needed for successful treatment. Moreover, the medical communication needs a new understanding that is out of empathy.

• Journal of Food Hygiene and Safety
• /
• v.36 no.1
• /
• pp.42-50
• /
• 2021

Along with the rapid growth of the food service industry, food safety requirements and hygiene are increasing in importance in restaurants and hotels. Accordingly, there is a need for quick and practical monitoring techniques to determine hygiene status in the field. In this study, we investigated 5 domestic 5-star hotels specifically, personal hygiene (hands of workers), cooking utensils (knife, cutting board, food storage container, slicing machine blade, ice-maker scoop) and other facilities (refrigerator handle, sink). In addition, we examined the hygiene management status of customer contact points (tongs for buffet, etc.) to derive the correlation between the ATP values as a, a verification method. As a result of our five-hotel survey, we found that cooking utensils and personal hygiene were relatively sanitary compared to other inspection items (cookware 92.2%, personal hygiene 91.4%, facilities and equipment 76.19%, customer contact items 88.6%). According to our ATP-based mothod, kitchen utensils (51 ± 45 RLU/25㎠) were relatively clean compared to other with facilities and equipment (167 ± 123 RLU/25㎠). In the present study, we also evaluated the usefulness of the ATP bioluminescence method for monitoring surface hygiene at hotel restaurants. After correlation analysis of surveillance of hygienic status points and ATP assay, most results showed negative and high correlation (-0.64--0.89). Our ATP assay (92 ± 67 RLU/25㎠) of each item after cleaning showed signigicantly reduced results compared to the ATP assay (1020 ± 1254 RLU/25㎠) for normal status, thereby indicating its suitability as a tool to verify the validity of cleaning. By our results, ATP bioluminescence could be used as an effective tool for visual numerical evaluation of invisible contaminants.

• Journal of Nutrition and Health
• /
• v.54 no.1
• /
• pp.23-38
• /
• 2021

Purpose: This study was undertaken to evaluate the health, nutritional status and metabolic syndrome risk according to the dietary pattern of adult single-person households, using information obtained from the Korea National Health and Nutrition Examination Survey (KNHANES). Methods: Data were collected from the 2013-2016 KNHANES, of adults aged 19-64 years, belonging to single-person households. Based on cluster analysis, the dietary patterns of subjects were classified into three groups. The dietary behavior factors, health-related factors, nutritional status, and prevalence of metabolic syndrome obtained from KNHANES questionnaires were compared according to the individual dietary pattern. The nutrient intake data of the subjects were calculated using the semi-food frequency questionnaire. Moreover, blood and physical measurement data of the subjects were analyzed to obtain the prevalence of metabolic syndromes. Results: The major dietary intakes of subjects were classified as 'Rice and kimchi', 'Mixed', and 'Milk·dairy products and fruits' patterns. Characteristics of subjects based on their dietary pattern, gender, age, and education level were significantly different. The 'Milk and fruits' pattern showed low frequency of skipping breakfast and eating out, and had higher intake of dietary supplements. Frequency of alcohol intake and smoking rates were highest in the 'Mixed' pattern. Maximum nutrient intake of fat, vitamin A, riboflavin, vitamin C, niacin, calcium, phosphorus, and potassium was obtained in the 'Milk·dairy products and fruits' pattern. According to dietary patterns adjusted for age and gender, the risk of metabolic syndrome was 0.380 times lower in the 'Milk·dairy products and fruit' pattern than in the 'Rice and kimchi' pattern. However, when adjusted for other confounding factors, no significant difference was obtained between dietary patterns for metabolic syndrome risk. Conclusion: These results indicate that the health and nutritional status of a single-person household is possibly affected by the dietary intake of subjects.