• Journal of the Korea Convergence Society
• /
• v.10 no.9
• /
• pp.23-29
• /
• 2019

In this paper, the mood of images was classified into eight categories through a deep convolutional neural network and video was automatically generated using proper background music. Based on the collected image data, the classification model is learned using a multilayer perceptron (MLP). Using the MLP, a video is generated by using multi-class classification to predict image mood to be used for video generation, and by matching pre-classified music. As a result of 10-fold cross-validation and result of experiments on actual images, each 72.4% of accuracy and 64% of confusion matrix accuracy was achieved. In the case of misclassification, by classifying video into a similar mood, it was confirmed that the music from the video had no great mismatch with images.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.9
• /
• pp.57-64
• /
• 2021

In this paper, we propose a device that can control the primary control surface of an aircraft by recognizing speech commands. The speech command consists of 19 commands, and a learning model is constructed based on a total of 2,500 datasets. The training model is composed of a CNN model using the Sequential library of the TensorFlow-based Keras model, and the speech file used for training uses the MFCC algorithm to extract features. The learning model consists of two convolution layers for feature recognition and Fully Connected Layer for classification consists of two dense layers. The accuracy of the validation dataset was 98.4%, and the performance evaluation of the test dataset showed an accuracy of 97.6%. In addition, it was confirmed that the operation was performed normally by designing and implementing a Raspberry Pi-based control device. In the future, it can be used as a virtual training environment in the field of voice recognition automatic flight and aviation maintenance.

• International Journal of Computer Science & Network Security
• /
• v.22 no.5
• /
• pp.294-302
• /
• 2022

Detection of fake news is a complex and a challenging task. Generation of fake news is very hard to stop, only steps to control its circulation may help in minimizing its impacts. Humans tend to believe in misleading false information. Researcher started with social media sites to categorize in terms of real or fake news. False information misleads any individual or an organization that may cause of big failure and any financial loss. Automatic system for detection of false information circulating on social media is an emerging area of research. It is gaining attention of both industry and academia since US presidential elections 2016. Fake news has negative and severe effects on individuals and organizations elongating its hostile effects on the society. Prediction of fake news in timely manner is important. This research focuses on detection of fake news spreaders. In this context, overall, 6 models are developed during this research, trained and tested with dataset of PAN 2020. Four approaches N-gram based; user statistics-based models are trained with different values of hyper parameters. Extensive grid search with cross validation is applied in each machine learning model. In N-gram based models, out of numerous machine learning models this research focused on better results yielding algorithms, assessed by deep reading of state-of-the-art related work in the field. For better accuracy, author aimed at developing models using Random Forest, Logistic Regression, SVM, and XGBoost. All four machine learning algorithms were trained with cross validated grid search hyper parameters. Advantages of this research over previous work is user statistics-based model and then ensemble learning model. Which were designed in a way to help classifying Twitter users as fake news spreader or not with highest reliability. User statistical model used 17 features, on the basis of which it categorized a Twitter user as malicious. New dataset based on predictions of machine learning models was constructed. And then Three techniques of simple mean, logistic regression and random forest in combination with ensemble model is applied. Logistic regression combined in ensemble model gave best training and testing results, achieving an accuracy of 72%.

• Journal of Korea Trade
• /
• v.24 no.8
• /
• pp.171-188
• /
• 2020

Purpose - This is an exploratory study that aims to apply text mining techniques, which computationally extracts words from the large-scale text data, to legal documents to quantify trade claim contents and enables statistical analysis. Design/methodology - This is designed to verify the validity of the application of text mining techniques as a quantitative methodology for trade claim studies, that have relied mainly on a qualitative approach. The subjects are 81 cases of arbitration and court judgments from China published on the website of the UNCITRAL where the CISG was applied. Validation is performed by comparing the manually analyzed result with the automatically analyzed result. The manual analysis result is the cluster analysis wherein the researcher reads and codes the case. The automatic analysis result is an analysis applying text mining techniques to the result of the cluster analysis. Topic modeling and semantic network analysis are applied for the statistical approach. Findings - Results show that the results of cluster analysis and text mining results are consistent with each other and the internal validity is confirmed. And the degree centrality of words that play a key role in the topic is high as the between centrality of words that are useful for grasping the topic and the eigenvector centrality of the important words in the topic is high. This indicates that text mining techniques can be applied to research on content analysis of trade claims for statistical analysis. Originality/value - Firstly, the validity of the text mining technique in the study of trade claim cases is confirmed. Prior studies on trade claims have relied on traditional approach. Secondly, this study has an originality in that it is an attempt to quantitatively study the trade claim cases, whereas prior trade claim cases were mainly studied via qualitative methods. Lastly, this study shows that the use of the text mining can lower the barrier for acquiring information from a large amount of digitalized text.

• Archives of Plastic Surgery
• /
• v.50 no.3
• /
• pp.254-263
• /
• 2023

Background The three-dimensional (3D) evaluation of skeletal stability after orthognathic surgery is a time-consuming and complex procedure. The complexity increases further when evaluating the surgery-first orthognathic approach (SFOA). Herein, we propose and validate a simple time-saving method of 3D analysis using a single software, demonstrating high accuracy and repeatability. Methods This retrospective cohort study included 12 patients with skeletal class 3 malocclusion who underwent bimaxillary surgery without any presurgical orthodontics. Computed tomography (CT)/cone-beam CT images of each patient were obtained at three different time points (preoperation [T0], immediately postoperation [T1], and 1 year after surgery [T2]) and reconstructed into 3D images. After automatic surface-based alignment of the three models based on the anterior cranial base, five easily located anatomical landmarks were defined to each model. A set of angular and linear measurements were automatically calculated and used to define the amount of movement (T1-T0) and the amount of relapse (T2-T1). To evaluate the reproducibility, two independent observers processed all the cases, One of them repeated the steps after 2 weeks to assess intraobserver variability. Intraclass correlation coefficients (ICCs) were calculated at a 95% confidence interval. Time required for evaluating each case was recorded. Results Both the intra- and interobserver variability showed high ICC values (more than 0.95) with low measurement variations (mean linear variations: 0.18 mm; mean angular variations: 0.25 degree). Time needed for the evaluation process ranged from 3 to 5 minutes. Conclusion This approach is time-saving, semiautomatic, and easy to learn and can be used to effectively evaluate stability after SFOA.

• Journal of Information Technology Services
• /
• v.22 no.4
• /
• pp.97-110
• /
• 2023

In this study, we propose a technique to automatically generate transfer documents using sensor data from livestock manure transfer systems. The research involves analyzing sensor data and applying machine learning techniques to derive optimized outcomes for livestock manure transfer documents. By comparing and contrasting with existing documents, we present a method for automatic document generation. Specifically, we propose the utilization of Gradient Boosting, a machine learning algorithm. The objective of this research is to enhance the efficiency of livestock manure and liquid byproduct management. Currently, stakeholders including producers, transporters, and processors manually input data into the livestock manure transfer management system during the disposal of manure and liquid byproducts. This manual process consumes additional labor, leads to data inconsistency, and complicates the management of distribution and treatment. Therefore, the aim of this study is to leverage data to automatically generate transfer documents, thereby increasing the efficiency of livestock manure and liquid byproduct management. By utilizing sensor data from livestock manure and liquid byproduct transport vehicles and employing machine learning algorithms, we establish a system that automates the validation of transfer documents, reducing the burden on producers, transporters, and processors. This efficient management system is anticipated to create a transparent environment for the distribution and treatment of livestock manure and liquid byproducts.

• Journal of the Institute of Convergence Signal Processing
• /
• v.23 no.2
• /
• pp.62-69
• /
• 2022

An electrocardiogram (ECG) is one of the most important biosignals, and in particular, continuous ECG monitoring is very important in patients with arrhythmia. There are many different types of arrhythmia (sinus node, sinus tachycardia, atrial premature beat (APB), and ventricular fibrillation) depending on the cause, and continuous ECG monitoring during daily life is very important for early diagnosis of arrhythmias and setting treatment directions. The ECG signal of arrhythmia patients is very unstable, and it is difficult to detect the R-peak point, which is a key feature for automatic arrhythmias detection. In this study, we develped a continuous measuring Holter ECG monitoring device and software for analysis and confirmed the utility of R-peak of the ECG signal with MIT-BIH arrhythmia database. In future studies, it needs the validation of algorithms and clinical data for morphological classification and prediction of arrhythmias due to various etiologies.

• Journal of Microbiology and Biotechnology
• /
• v.33 no.10
• /
• pp.1337-1350
• /
• 2023

Caproic acid is a precursor substance for the synthesis of ethyl caproate, the main flavor substance of nongxiangxing baijiu liquor. In this study, Clostridium butyricum GD1-1, a strain with high caproic acid concentration (3.86 g/l), was isolated from the storage pit mud of nongxiangxing baijiu for sequencing and analysis. The strain's genome was 3,840,048 bp in length with 4,050 open reading frames. In addition, virulence factor annotation analysis showed C. butyricum GD1-1 to be safe at the genetic level. However, the annotation results using the Kyoto Encyclopedia of Genes and Genomes Automatic Annotation Server predicted a deficiency in the strain's synthesis of alanine, methionine, and biotin. These results were confirmed by essential nutrient factor validation experiments. Furthermore, the optimized medium conditions for caproic acid concentration by strain GD1-1 were (g/l): glucose 30, NaCl 5, yeast extract 10, peptone 10, beef paste 10, sodium acetate 11, L-cysteine 0.6, biotin 0.004, starch 2, and 2.0% ethanol. The optimized fermentation conditions for caproic acid production by C. butyricum GD1-1 on a single-factor basis were: 5% inoculum volume, 35℃, pH 7, and 90% loading volume. Under optimal conditions, the caproic acid concentration of strain GD1-1 reached 5.42 g/l, which was 1.40 times higher than the initial concentration. C. butyricum GD1-1 could be further used in caproic acid production, NXXB pit mud strengthening and maintenance, and artificial pit mud preparation.

• International Journal of Computer Science & Network Security
• /
• v.24 no.3
• /
• pp.125-134
• /
• 2024

The novel coronavirus 2019 is called COVID-19 has outspread swiftly worldwide. An early diagnosis is more important to control its quick spread. Medical imaging mechanics, chest calculated tomography or chest X-ray, are playing a vital character in the identification and testing of COVID-19 in this present epidemic. Chest X-ray is cost effective method for Covid-19 detection however the manual process of x-ray analysis is time consuming given that the number of infected individuals keep growing rapidly. For this reason, it is very important to develop an automated COVID-19 detection process to control this pandemic. In this study, we address the task of automatic detection of Covid-19 by using a popular deep learning model namely the VGG19 model. We used 1300 healthy and 1300 confirmed COVID-19 chest X-ray images in this experiment. We performed three experiments by freezing different blocks and layers of VGG19 and finally, we used a machine learning classifier SVM for detecting COVID-19. In every experiment, we used a five-fold cross-validation method to train and validated the model and finally achieved 98.1% overall classification accuracy. Experimental results show that our proposed method using the deep learning-based VGG19 model can be used as a tool to aid radiologists and play a crucial role in the timely diagnosis of Covid-19.

• Korean Journal of Radiology
• /
• v.22 no.3
• /
• pp.405-414
• /
• 2021

Objective: To compare two clinically available MR volumetry software, NeuroQuant^® (NQ) and Inbrain^® (IB), and examine the inter-method reliabilities and differences between them. Materials and Methods: This study included 172 subjects (age range, 55-88 years; mean age, 71.2 years), comprising 45 normal healthy subjects, 85 patients with mild cognitive impairment, and 42 patients with Alzheimer's disease. Magnetic resonance imaging scans were analyzed with IB and NQ. Mean differences were compared with the paired t test. Inter-method reliability was evaluated with Pearson's correlation coefficients and intraclass correlation coefficients (ICCs). Effect sizes were also obtained to document the standardized mean differences. Results: The paired t test showed significant volume differences in most regions except for the amygdala between the two methods. Nevertheless, inter-method measurements between IB and NQ showed good to excellent reliability (0.72 < r < 0.96, 0.83 < ICC < 0.98) except for the pallidum, which showed poor reliability (left: r = 0.03, ICC = 0.06; right: r = -0.05, ICC = -0.09). For the measurements of effect size, volume differences were large in most regions (0.05 < r < 6.15). The effect size was the largest in the pallidum and smallest in the cerebellum. Conclusion: Comparisons between IB and NQ showed significantly different volume measurements with large effect sizes. However, they showed good to excellent inter-method reliability in volumetric measurements for all brain regions, with the exception of the pallidum. Clinicians using these commercial software should take into consideration that different volume measurements could be obtained depending on the software used.