- P-ISSN1738-6764
- E-ISSN2093-7504
- KCI
ISSN : 1738-6764
Vol.18 No.4
Abstract
In recent decades, anomaly detection has undoubtedly become one of the most important areas of research. This is because applications such as financial transactions, medical fraud, and anomaly detection can be used to solve a wide range of real-life problems. Data from these applications can be modeled using large graphs of many different nodes and edges. Because of the size and heterogeneity of the data contained in the graph, it is a very difficult task to detect abnormal patterns. In this paper, we proposed a method for detecting abnormal patterns in a large homogeneous graph. The proposed method consisted of two steps. In the first step, the graph was transformed into a vector using a semi-supervised graph neural network (GCN). The second step was based on DBSCAN, an unsupervised clustering method. Various performance evaluations were performed to show the superiority of the proposed method. Experimental results showed that the proposed method could detect abnormal nodes with high accuracy in homogeneous static graphs.
Abstract
We examined how knowledge sharing behavior on social network sites is associated with impression management motivations as knowledge sharing can be regarded as a self-representation behavior that people perform to impress others. We collected data through an online survey in which 705 respondents participated. To answer the research questions and test hypotheses, a set of hierarchical regression models were employed. We found that both motivation for impression management and the explanatory variables of self-determination theory played important roles in explaining knowledge sharing behavior on Facebook. However, the explanatory variables of self-determination theory were positively associated with the dependent variable, whereas motivations for impression management were negatively associated. We further found that the exact relationship between the motivation to impress others and knowledge sharing varied according to the topic of the knowledge that people shared on Facebook. That is, impression management motivation played a more negative role for topics that require more expertise, such as “politics,” “economics,” and “health” topics, than for more casual topics. This suggests that people might have greater fear of negative evaluation when they are considering posts related to “politics,” “economics,” and “health” topics.
Abstract
The purpose of this study was to investigate whether the high pitch protocol method of dual-energy computed tomography can be successfully applied to a lumbar vertebrae CT scan and reduce the effect of exposure dose; the main indicators were the image quality and reduced scan time. This study was conducted using MDCT 256 row dual source, Somatom Definition Flash (Siemens medical system: Germany), on 69 patients who had lumbar vertebrae CT scans from March 1, 2021, to June 30, 2021. First, to determine the applicability of the high pitch protocol method, the phantom images quality was compared with that of the existing lumbar vertebrae CT scan method and the examination method using the high pitch protocol method with the AAPM phantom, respectively. In the phantom image, CT noise, uniformity, and spatial resolution were not significantly different between the existing lumbar vertebrae CT scan method and the examination method using the high pitch protocol. In addition, in images examined from clinical patients, single-sample ttest results of the comparison of the mean value of CT noise, uniformity, and spatial resolution did not show statistically significant differences, 0.081, 0.078, and 0.387, respectively. The results of the qualitative evaluation showed that all were also higher than 3 points, which is fair, and the radiation dose was reduced by 9.76% in the high pitch protocol. When examining the lumbar region with dual-energy computed tomography, the examination method using high-pitch mode reduced the radiation dose and shortened the examination time, with the same image quality as that of the existing lumbar vertebrae CT scan.
Abstract
This study analyzes the changes in the efficiency of Korean R&D projects in the IT sector from 2010 to 2018. The data envelopment analysis (DEA) and Cumulative Malmquist index were used to measure the change in R&D efficiency. The results were as follows. First, the Korean IT sector has developed around hardware, applied research, development research, and private enterprises rather than software, basic research, universities, and public research institutes. Second, short-term efficiency improvement is possible in the government's R&D projects in which large-scale investments are executed. Third, in Korea, it is more urgent to improve efficiency in the public sector than in the private sector. The results provide insights into policy implications required to enhance national competitiveness in the era of the 4th industrial revolution and digital transformation.
Abstract
The purpose of this paper was to evaluate the usability of Korean logistics in life (LIL) startups mobile apps that aim to provide more convenience and a better quality of life. Determining an app’s usability requires a different set of evaluation criteria from that of desktop sites or full sites due to the inherent characteristics of the mobile environment. This paper derived 10 evaluation criteria from a literature review, calculated the weights of each criterion, and performed a usability test for 14 Korean LIL apps involving 27 subjects. To highlight areas of improvement, a comparison was done using a top ranked app on Google Play which had the best score in terms of usability (in 10 items test) among the best Korean LIL app. The average usability score of the LIL apps represented an overall lack of high usability. The evaluation criteria of the apps showed higher scores in the error-free, prediction, and consistency variables categories. In contrast, this was not the case for effectiveness, usefulness, and interaction, which were relatively low. Usability and interest had a slightly negative relationship when implementing big data analytics. From the quadrants analysis, it was found that for most of the evaluation criteria, continuous improvement was needed. As a result of logistics 4.0, the number of LIL apps is growing rapidly. This paper is the first to define the concept of LIL and develop an evaluation model for usability, which was then used to identify the ease of use, learnability, effectiveness, and efficiency of representative LIL apps in Korea.
Abstract
This study aimed to explore a Korean medical student’s perception of professional and unprofessional behaviors by peers and physicians. A total of 144(56.92%) out of 253 graduating students from three medical schools in Korea completed an online survey. The survey used the Climate of Professionalism in a Clinical Teaching Environment scale, which queries the subjects’ observations of professional and unprofessional behaviours. An analysis was derived from reliability, factor analysis, descriptive statistics, and a paired t-test. The reliability was identified in the peers (Cronbach’s alpha=.83) and the physicians (Cronbach’s alpha= .88). By the respondents’ reports, both the peers and the physicians were less professional in their error handling. The difference in observed behaviours between the two groups existed in “disrespect to others” (p< .000), “the expense of others” (p< .05), and “ignorance of the other’s unprofessional behaviour” (p< .000). In their clinical practice, the students are watching professional or unprofessional behaviours of peer students and physicians. Instructional intervention is needed to ensure that the unprofessional behaviours observed by the students are not transformed into a learning experience.
Abstract
Vision represents an individual’s future image and is related to one’s psychological state. An individual’s second-language (L2) vision determines their language study goals, and religious or spiritual beliefs may be central to this vision of oneself. This study aimed to discover the participants’ L2 visions and explore how the Christian faith affects their L2 visions. The researchers adopted a qualitative research approach to examine how Christianity informed the L2 visions and English learning of four Chinese Christian graduate students. This study was performed in a provincial university in China and lasted from March 2021 to December 2021. Data were collected through interviews, observations, and informal talk. The collected data were analyzed using thematic analysis containing initial and axial coding. The findings showed that the participants’ L2 visions included their dissatisfaction with the L2 reality and the identification of their L2 expectations. Moreover, the data revealed that the participants’ Christian faith brought them devotion and self-discipline in English learning. Finally, the researchers found that the L2 vision and Christian faith can be combined into a unique vision to stimulate students’ English learning. One pedagogical implication is that students need to be encouraged by their teachers to discover their visions and develop different ones. Another implication is that teachers should simultaneously pay attention to students’ visions and adjust their teaching strategies to meet students’ needs.
Abstract
This study introduced an analytical framework for assessing the inter-organizational network system and the web impacts of public diplomacy organizations’ sites through the cases of South Korea and China. This study compared interorganizational collaboration networks, the impact of the government agencies’ websites, and the sectoral and geographic distribution of information resources offered by the agencies on the web. Social network analysis was employed, and it indicated that the Chinese public diplomacy organizations constructed denser and more strongly connected networks than the Korean public diplomacy agencies. Furthermore, the results suggested that .com was the most popular generic top-level domain, followed by .org, .net, and .edu, for both Korean and Chinese organizations. The source sites that sent links to Korean organizations originated mostly from East Asian (Korea, Indonesia, and Japan) and European countries (Germany and Russia). Information about Chinese culture was spread more widely across diverse countries, including East Asian (China and Japan), North American (Canada), European (United Kingdom, France), and Oceanian (Australia) countries. For both Korea and China, domestic audiences played key roles as information hubs in each network, which illuminates a networked and cooperative form of digital diplomacy outreach in these countries.
Abstract
This study aimed to shed light on the determinants of purchasing intention in direct-to-consumer (D2C) online stores in South Korea in the context of digital transformation and the COVID-19 pandemic. The study recruited 315 Korean consumers who had experience shopping at D2C e-commerce stores to respond. The survey items were: brand coolness traits, brand attitude, service qualities of D2C online stores, and price consciousness. The survey revealed that three of the four brand coolness traits (high status, popular, iconic) had a positive influence on brand attitude. The results also showed that brand attitude, personalized service, and product assortment size had significant impacts on purchasing intention. However, energetic, which is one of the four brand coolness traits and price consciousness did not have an influence on purchasing intention at a significant level. The current findings will help marketers understand consumers’ preferences in using D2C e-commerce stores as opposed to traditional e-commerce stores.
Abstract
Among various fields at home and abroad, machine learning technology is developing rapidly, centering on deep learning. Active research is being conducted. However, due to the lack of sufficient machine learning data, opportunities such as providing high value-added content and improving service performance are missed. New entity names are registered in various documents and interpreted as different entity names depending on the context. Through prescriptive analysis-based key entity name recognition technology that collects researcher's research field, research propensity, and research activity information and converts the value of data into new phrase-unit information into word-unit qualities by constructing and utilizing generalpurpose data, we provide a research information curation service based on prescriptive analysis to solve difficulties of securing learning data for researchers who want to apply machine learning and for obtaining customized research information that corresponds to individualized services for each researcher.
Abstract
When training neural networks with softmax outputs, one-hot target encoding is commonly used due to its simplicity. This strategy does not incorporate the probability that an input sample belongs to a certain class but adopts “one” or “zero” as the desired output values of neural networks. Instead of the most prevalent one-hot encoding, this paper proposes a probabilistic target encoding to prevent the overfitting of neural networks to training samples. This effect brings about the accuracy improvement of test samples. We demonstrated the effectiveness of the proposed target encoding through simulations of multi-layer perceptrons and convolutional neural networks for various classification problems such as handwritten-digit recognition, isolated-word recognition, image classification, and object recognition tasks. The simulation results show that the proposed probabilistic target encoding is superior to one-hot encoding as it prevents the overfitting of neural networks to training samples.
Abstract
As interest in the development of artificial intelligence technology in the water supply field increases, an artificial neural network algorithm that can predict improved decision ratings through repetitive learning using results of aging pipe condition evaluation data should be developed and the most reliable prediction model should be presented through a verification process. An algorithm was developed to predict pipeline ratings by updating weights through backpropagation so that 12 items of indirect evaluation data according to the 2020 Han River Basin's basic plan could be pre-processed, such as standardizing input values and applying artificial neural network algorithms. As a result of algorithm accuracy verification, if there is sufficient data and the repetitive learning and upgrades are continuously conducted, the prediction accuracy will become higher and a reliable AI-based water supply pipe condition evaluation model to be used nationwide will be developed in the future.
Abstract
The objective of this study was to design contents of Technology Enhanced Learning (TEL) with emphasis on a co-curricular program to promote effective operation of an online-based curriculum. The program with TEL emphasis enhanced operations and strengthened core-competencies in higher education. To achieve this goal, we examined literature reviews and case studies of the co-curricular program with a focus on the theoretical foundation of the digital instructional design model. Our cases derived from T University located in Busan, South Korea were used to understand how the co-curricular program was operated and a template of an online-based class model was designed to perform in both online and offline classrooms. Results of this study are expected to further the exploration of co-curricular learning model as well as cement the embellishment of institutional strategies for technology-enhanced digital teaching and learning methods.