• Smart Media Journal
• /
• v.12 no.1
• /
• pp.9-16
• /
• 2023

Recently, a convolutional neural network (CNN) based system is being developed to overcome the limitations of human resources in the apple quality classification of farmhouse. However, since convolutional neural networks receive only images of the same size, preprocessing such as sampling may be required, and in the case of oversampling, information loss of the original image such as image quality degradation and blurring occurs. In this paper, in order to minimize the above problem, to generate a image patch based graph of an original image and propose a random walk-based positional encoding method to apply the graph transformer model. The above method continuously learns the position embedding information of patches which don't have a positional information based on the random walk algorithm, and finds the optimal graph structure by aggregating useful node information through the self-attention technique of graph transformer model. Therefore, it is robust and shows good performance even in a new graph structure of random node order and an arbitrary graph structure according to the location of an object in an image. As a result, when experimented with 5 apple quality datasets, the learning accuracy was higher than other GNN models by a minimum of 1.3% to a maximum of 4.7%, and the number of parameters was 3.59M, which was about 15% less than the 23.52M of the ResNet18 model. Therefore, it shows fast reasoning speed according to the reduction of the amount of computation and proves the effect.

• Convergence Security Journal
• /
• v.23 no.2
• /
• pp.47-54
• /
• 2023

The global popularity of K-content(Korean Wave) has led to a continuous increase in copyright infringement cases involving domestic works, not only within the country but also overseas. In response to this trend, there is active research on technologies for detecting illegal distribution sites of domestic copyrighted materials, with recent studies utilizing the characteristics of domestic illegal distribution sites that often include a significant number of advertising banners. However, the application of detection techniques similar to those used domestically is limited for overseas illegal distribution sites. These sites may not include advertising banners or may have significantly fewer ads compared to domestic sites, making the application of detection technologies used domestically challenging. In this study, we propose a detection technique based on the similarity comparison of links and text trees, leveraging the characteristic of including illegal sharing posts and images of copyrighted materials in a similar hierarchical structure. Additionally, to accurately compare the similarity of large-scale trees composed of a massive number of links, we utilize Graph Neural Network (GNN). The experiments conducted in this study demonstrated a high accuracy rate of over 95% in classifying regular sites and sites involved in the illegal distribution of copyrighted materials. Applying this algorithm to automate the detection of illegal distribution sites is expected to enable swift responses to copyright infringements.

• International conference on construction engineering and project management
• /
• 2011.02a
• /
• pp.432-438
• /
• 2011

The purpose of the present thesis is to develop bankruptcy prediction models capable of being applied to the Korean construction industry and to deduce an optimal model through comparative evaluation of final developed models. A study population was selected as general contractors in the Korean construction industry. In order to ease the sample securing and reliability of data, it was limited to general contractors receiving external audit from the government. The study samples are divided into a bankrupt company group and a non-bankrupt company group. The bankruptcy, insolvency, declaration of insolvency, workout and corporate reorganization were used as selection criteria of a bankrupt company. A company that is not included in the selection criteria of the bankrupt company group was selected as a non-bankrupt company. Accordingly, the study sample is composed of a total of 112 samples and is composed of 48 bankrupt companies and 64 non-bankrupt companies. A financial ratio was used as early predictors for development of an estimation model. A total of 90 financial ratios were used and were divided into growth, profitability, productivity and added value. The MDA (Multivariate Discriminant Analysis) model and BLRA (Binary Logistic Regression Analysis) model were used for development of bankruptcy prediction models. The MDA model is an analysis method often used in the past bankruptcy prediction literature, and the BLRA is an analysis method capable of avoiding equal variance assumption. The stepwise (MDA) and forward stepwise method (BLRA) were used for selection of predictor variables in case of model construction. Twenty two variables were finally used in MDA and BLRA models according to timing of bankruptcy. The ROC-Curve Analysis and Classification Analysis were used for analysis of prediction performance of estimation models. The correct classification rate of an individual bankruptcy prediction model is as follows: 1) one year ago before the event of bankruptcy (MDA: 83.04%, BLRA: 93.75%); 2) two years ago before the event of bankruptcy (MDA: 77.68%, BLRA: 78.57%); 3) 3 years ago before the event of bankruptcy (MDA: 84.82%, BLRA: 91.96%). The AUC (Area Under Curve) of an individual bankruptcy prediction model is as follows. : 1) one year ago before the event of bankruptcy (MDA: 0.933, BLRA: 0.978); 2) two years ago before the event of bankruptcy (MDA: 0.852, BLRA: 0.875); 3) 3 years ago before the event of bankruptcy (MDA: 0.938, BLRA: 0.975). As a result of the present research, accuracy of the BLRA model is higher than the MDA model and its prediction performance is improved.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.34 no.2
• /
• pp.207-216
• /
• 2024

As big data was built due to the 4th Industrial Revolution, personalized services increased rapidly. As a result, the amount of personal information collected from online services has increased, and concerns about users' personal information leakage and privacy infringement have increased. Online service providers provide privacy policies to address concerns about privacy infringement of users, but privacy policies are often misused due to the long and complex problem that it is difficult for users to directly identify risk items. Therefore, there is a need for a method that can automatically check whether the privacy policy is safe. However, the safety verification technique of the conventional blacklist and machine learning-based privacy policy has a problem that is difficult to expand or has low accessibility. In this paper, to solve the problem, we propose a safety verification technique for the privacy policy using the GPT-3.5 API, which is a generative artificial intelligence. Classification work can be performed evenin a new environment, and it shows the possibility that the general public without expertise can easily inspect the privacy policy. In the experiment, how accurately the blacklist-based privacy policy and the GPT-based privacy policy classify safe and unsafe sentences and the time spent on classification was measured. According to the experimental results, the proposed technique showed 10.34% higher accuracy on average than the conventional blacklist-based sentence safety verification technique.

• Journal of Sasang Constitutional Medicine
• /
• v.18 no.2
• /
• pp.41-49
• /
• 2006

1. Objectives The Face is an important standard for the classification of Sasang Contitutions. We are developing 3D Face Automatic Recognition Apparatus to analyse the facial characteristics. So We should examine a shape demobilization error of 3D Face Automatic Recognition Apparatus. 2. Methods We compared facial shape data be demobilized by 3D Face Automatic Recognition Apparatus with facial shape data that be demobilized by 3D laser scanner. The subject was two korean men. And We analysed the average error and the maximum error of two data. In this process, We used one datum point(the peak of nose) and two datum line(vertical section and horizontal section). 3. Results and Conclusions In each this comparison, the average error of vertical section was 1.962574mm and 2.703814mm. and the maximum error of vertical section was 16.968249mm and 18.61464mm. the average error of horizontal section was 4.173203mm and 21.487479mm. and the maximum error of horizontal section was 3.571210mm and 17.13255mm. Also We complemented this apparatus a little and We reexamined a shape demobilization error of 3D Face Automatic Recognition Apparatus again. Accuracy of a shape demobilization was improved a little. From now on We complement accuracy of a shape demobilization in 3D Face Recognition Apparatus.

• Journal of the Korea Convergence Society
• /
• v.11 no.8
• /
• pp.99-106
• /
• 2020

The purpose of this study is to find out whether image similarity between celebrities and brands on social network service be a determinant to predict advertising effectiveness. To this end, an advertising effect prediction model for celebrity endorsed advertising was created and its validity was verified through a machine learning method which is a big data analysis technique. Firstly, the celebrity-brand image similarity, which was used as an independent variable, was quantified by the association network theory with social big data, and secondly a multiple regression model which used data representing advertising effects as a dependent variable was repeatedly conducted to generate an advertising effect prediction model. The accuracy of the prediction model was decided by comparing the prediction results with the survey outcomes. As for a result, it was proved that the validity of the predictive modeling of advertising effects was secured since the classification accuracy of 75%, which is a criterion for judging validity, was shown. This study suggested a new methodological alternative and direction for big data-based modeling research through celebrity-brand image similarity structure based on social network theory, and effect prediction modeling by machine learning.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.11
• /
• pp.1-8
• /
• 2020

In recent years, due to the exponential increase in internet data, many fields such as deep learning have developed, but side effects generated as commercial advertisements, such as viral marketing, have been discovered. This not only damages the essence of the internet for sharing high-quality information, but also causes problems that increase users' search times to acquire high-quality information. In this study, we define advertisement as "a text that obscures the essence of information transmission" and we propose a model for filtering information according to that definition. The proposed model consists of advertisement filtering and advertisement filtering performance improvement and is designed to continuously improve performance. We collected data for filtering advertisements and learned document classification using KorBERT. Experiments were conducted to verify the performance of this model. For data combining five topics, accuracy and precision were 89.2% and 84.3%, respectively. High performance was confirmed, even if atypical characteristics of advertisements are considered. This approach is expected to reduce wasted time and fatigue in searching for information, because our model effectively delivers high-quality information to users through a process of determining and filtering advertisement paragraphs.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.28 no.2
• /
• pp.385-395
• /
• 2018

With the development of Internet, cyber attack has become a major threat. To detect cyber attacks, intrusion detection system(IDS) has been widely deployed. But IDS has a critical weakness which is that it generates a large number of false alarms. One of the promising techniques that reduce the false alarms in real time is machine learning. However, there are problems that must be solved to use machine learning. So, many machine learning approaches have been applied to this field. But so far, researchers have not focused on features. Despite the features of IDS alerts are important for performance of model, the approach to feature is ignored. In this paper, we propose new feature set which can improve the performance of model and can be extracted from a single alarm. New features are motivated from security analyst's know-how. We trained and tested the proposed model applied new feature set with real IDS alerts. Experimental results indicate the proposed model can achieve better accuracy and false positive rate than SVM model with ordinary features.

• Atmosphere
• /
• v.24 no.3
• /
• pp.433-444
• /
• 2014

Although the Radar-AWS Rainrate (RAR) calculation system operated by Korea Meteorological Administration estimated precipitation using 2-dimensional composite components of single polarization radars, this system has several limitations in estimating the precipitation accurately. To to overcome limitations of the RAR system, the Korea Meteorological Administration developed and operated the RMQ (Radar-based Multi-sensor Quantitative Precipitation Estimation) system, the improved version of NMQ (National Mosaic and Multi-sensor Quantitative Precipitation Estimation) system of NSSL (National Severe Storms Laboratory) for the Korean Peninsula. This study introduced the RMQ system domestically for the first time and verified the precipitation estimation performance of the RMQ system. The RMQ system consists of 4 main parts as the process of handling the single radar data, merging 3D reflectivity, QPE, and displaying result images. The first process (handling of the single radar data) has the pre-process of a radar data (transformation of data format and quality control), the production of a vertical profile of reflectivity and the correction of bright-band, and the conduction of hydrid scan reflectivity. The next process (merger of 3D reflectivity) produces the 3D composite reflectivity field after correcting the quality controlled single radar reflectivity. The QPE process classifies the precipitation types using multi-sensor information and estimates quantitative precipitation using several Z-R relationships which are proper for precipitation types. This process also corrects the precipitation using the AWS position with local gauge correction technique. The last process displays the final results transformed into images in the web-site. This study also estimated the accuracy of the RMQ system with five events in 2012 summer season and compared the results of the RAR (Radar-AWS Rainrate) and RMQ systems. The RMQ system ($2.36mm\;hr^{-1}$ in RMSE on average) is superior to the RAR system ($8.33mm\;hr^{-1}$ in RMSE) and improved by 73.25% in RMSE and 25.56% in correlation coefficient on average. The precipitation composite field images produced by the RMQ system are almost identical to the AWS (Automatic Weather Statioin) images. Therefore, the RMQ system has contributed to improve the accuracy of precipitation estimation using weather radars and operation of the RMQ system in the work field in future enables to cope with the extreme weather conditions actively.

• Analytical Science and Technology
• /
• v.28 no.5
• /
• pp.309-316
• /
• 2015

The analytical method of lead in plasma by ICP-MS was validated after securing environment within class 1,000 classification. We tested specificity and accuracy of within-run and between-run. According to measurement of the amount of suspended particulates in a clean room, 0.3~62 particles were detected in 0.3 µm size while 0.0~28.3 particles were observed in 0.5 µm size. Total suspended particulates met required environment with up to 90.3 particles. The MDL (Method detection limit) of the sample which has been fabricated using fetal bovine serum (FBS) blank was 1.77 ng/L, and LOQ (Limit of quantification) was 5.55 ng/L. The slope, intercept and correlation coefficient of the calibration curve were y=1.09×10⁻³x+4.88×10⁻² and r=0.9999, which showed good correlation. The specificity, within-run and between-run accuracy satisfied the standard at more than 50 ng/L. The average lead concentration in plasma of the general people, current workers and retired workers was 55.4 ng/L, 440 ng/L, and 132 ng/L.