• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.20 no.5
• /
• pp.100-112
• /
• 2021

With the advent of big data, traffic prediction has been developed based on historical data analysis methods, but this method deteriorates prediction performance when a traffic incident that has not been observed occurs. This study proposes a method that can compensate for the reduction in traffic prediction accuracy in traffic incidents situations by hybrid approach of machine learning and traffic simulation. The blind spots of the data-driven method are revealed when data patterns that have not been observed in the past are recognized. In this study, we tried to solve the problem by reinforcing historical data using traffic simulation. The proposed method performs machine learning-based traffic prediction and periodically compares the prediction result with real time traffic data to determine whether an incident occurs. When an incident is recognized, prediction is performed using the synthetic traffic data generated through simulation. The method proposed in this study was tested on an actual road section, and as a result of the experiment, it was confirmed that the error in predicting traffic state in incident situations was significantly reduced. The proposed traffic prediction method is expected to become a cornerstone for the advancement of traffic prediction.

• The Journal of Bigdata
• /
• v.6 no.1
• /
• pp.51-62
• /
• 2021

In the field of medical artificial intelligence, there have been a lot of researches on disease prediction and classification algorithms that can help doctors judge, but relatively less interested in artificial intelligence that can help medical consumers acquire and judge information. The fact that more than 150,000 questions have been asked about which hospital to go over the past year in NAVER portal will be a testament to the need to provide medical information suitable for medical consumers. Therefore, in this study, we wanted to establish a classification model that classifies 8 medical subjects for symptom text directly described by patients which was collected from NAVER portal to help consumers choose appropriate medical subjects for their symptoms. In order to ensure the validity of the data involving patients' subject matter, we conducted similarity measurements between objective symptom text (typical symptoms by medical subjects organized by the Seoul Emergency Medical Information Center) and subjective symptoms (NAVER data). Similarity measurements demonstrated that if the two texts were symptoms of the same medical subject, they had relatively higher similarity than symptomatic texts from different medical subjects. Following the above procedure, the classification model was constructed using a ridge regression model for subjective symptom text that obtained validity, resulting in an accuracy of 0.73.

• Korean Journal of Remote Sensing
• /
• v.37 no.5_1
• /
• pp.1013-1027
• /
• 2021

Images taken with KOMPSAT-3 have additional NIR and PAN bands, as well as RGB regions of the visible ray band, compared to imagestaken with a standard camera. Furthermore, electrical and optical properties must be considered because a wide radius area of approximately 17 km or more is photographed at an altitude of 685 km above the ground. In other words, the camera sensor of KOMPSAT-3 is distorted by each CCD pixel, characteristics of each band,sensitivity and time-dependent change, CCD geometry. In order to solve the distortion, correction of the sensors is essential. In this paper, we propose a method for detecting uniform regions in side-slider-based KOMPSAT-3 images using segment-based noise analysis. After detecting a uniform area with the corresponding algorithm, a correction table was created for each sensor to apply the non-uniformity correction algorithm, and satellite image correction was performed using the created correction table. As a result, the proposed method reduced the distortion of the satellite image,such as vertical noise, compared to the conventional method. The relative radiation accuracy index, which is an index based on mean square error (RA) and an index based on absolute error (RE), wasfound to have a comparative advantage of 0.3 percent and 0.15 percent, respectively, over the conventional method.

• Convergence Security Journal
• /
• v.21 no.2
• /
• pp.57-66
• /
• 2021

Deep neural networks (DNNs) provide excellent performance for image, speech, and pattern recognition. However, DNNs sometimes misrecognize certain adversarial examples. An adversarial example is a sample that adds optimized noise to the original data, which makes the DNN erroneously misclassified, although there is nothing wrong with the human eye. Therefore studies on defense against adversarial example attacks are required. In this paper, we have experimentally analyzed the success rate of detection for adversarial examples by adjusting various parameters. The performance of the ensemble defense method was analyzed using fast gradient sign method, DeepFool method, Carlini & Wanger method, which are adversarial example attack methods. Moreover, we used MNIST as experimental data and Tensorflow as a machine learning library. As an experimental method, we carried out performance analysis based on three adversarial example attack methods, threshold, number of models, and random noise. As a result, when there were 7 models and a threshold of 1, the detection rate for adversarial example is 98.3%, and the accuracy of 99.2% of the original sample is maintained.

• Journal of the Korean Institute of Gas
• /
• v.25 no.2
• /
• pp.13-21
• /
• 2021

Because of the recent expansion of hydrogen vehicle supply, the installation of hydrogen filling station is expected to gradually expand. This study attempts to predict the damage scale and propose a safer design form based on the scenario that assumes the worst case of a hydrogen station. A Flacs solver using computational fluid dynamics (CFD) was used to predict the damage scale, and the accuracy was verified by comparing it with the experimental results of previous researchers. The damage scale prediction was conducted for hydrogen leakage and explosion, and the prediction target was the KR model based on the measured values. And as a comparative review model, a roofless model was selected without a ceiling. As a result of analyzing the two models, it was possible to confirm the accumulation and retention of hydrogen gas up to 60 vol% or more in the KR model, whereas in the case of the Roofless model, the phenomenon of discharge and diffusion to the outside of the charging station by riding the wall after leakage. I was able to check. In conclusion, it was reviewed that the type of hydrogen charging station without ceiling is more advantageous for safety than the hydrogen filling station model.

• Journal of the Korean Society of Radiology
• /
• v.15 no.2
• /
• pp.153-158
• /
• 2021

X-ray image analysis is a very important field to improve the early diagnosis rate and prediction accuracy of periodontal disease. Research on the development and application of artificial intelligence-based algorithms to improve the quality of such dental X-ray images is being widely conducted worldwide. Thus, the aim of this study was to design a super-resolution algorithm for predicting periodontal disease and to evaluate its applicability in dental X-ray images. The super-resolution algorithm was constructed based on the convolution layer and ReLU, and an image obtained by up-sampling a low-resolution image by 2 times was used as an input data. Also, 1,500 dental X-ray data used for deep learning training were used. Quantitative evaluation of images used root mean square error and structural similarity, which are factors that can measure similarity through comparison of two images. In addition, the recently developed no-reference based natural image quality evaluator and blind/referenceless image spatial quality evaluator were additionally analyzed. According to the results, we confirmed that the average similarity and no-reference-based evaluation values were improved by 1.86 and 2.14 times, respectively, compared to the existing bicubic-based upsampling method when the proposed method was used. In conclusion, the super-resolution algorithm for predicting periodontal disease proved useful in dental X-ray images, and it is expected to be highly applicable in various fields in the future.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_3
• /
• pp.1965-1974
• /
• 2021

Satellite remote sensing approach can be actively used for forest monitoring. Especially, it is much meaningful to utilize Korea multi-purpose satellites, an independently operated satellite in Korea, for forest monitoring of Korea, Recently, several studies have been performed to exploit meaningful information from satellite remote sensed data via machine learning approaches. The forest information produced through machine learning approaches can be used to support the efficiency of traditional forest monitoring methods, such as in-situ survey or qualitative analysis of aerial image. The performance of machine learning approaches is greatly depending on the characteristics of study area and data. Thus, it is very important to survey the best model among the various machine learning models. In this study, the performance of deep neural network to classify artificial or natural forests was analyzed in Samcheok, Korea. As a result, the pixel accuracy was about 0.857. F1 scores for natural and artificial forests were about 0.917 and 0.433 respectively. The F1 score of artificial forest was low. However, we can find that the artificial and natural forest classification performance improvement of about 0.06 and 0.10 in F1 scores, compared to the results from single layered sigmoid artificial neural network. Based on these results, it is necessary to find a more appropriate model for the forest type classification by applying additional models based on a convolutional neural network.

• Journal of the Korea Convergence Society
• /
• v.12 no.12
• /
• pp.93-106
• /
• 2021

Blenderbot 2.0 is a dialogue model representing open domain chatbots by reflecting real-time information and remembering user information for a long time through an internet search module and multi-session. Nevertheless, the model still has many improvements. Therefore, this paper analyzes the limitations and errors of BlenderBot 2.0 from three perspectives: model, data, and dialogue. From the data point of view, we point out errors that the guidelines provided to workers during the crowdsourcing process were not clear, and the process of refining hate speech in the collected data and verifying the accuracy of internet-based information was lacking. Finally, from the viewpoint of dialogue, nine types of problems found during conversation and their causes are thoroughly analyzed. Furthermore, practical improvement methods are proposed for each point of view, and we discuss several potential future research directions.

• Journal of Korean Society of Forest Science
• /
• v.110 no.4
• /
• pp.610-621
• /
• 2021

The soil creep, primarily caused by earthquakes and torrential rainfall events, has widely occurred across the country. The Korea Forest Service attempted to quantify the soil creep susceptible areas using a discriminant value table to prevent or mitigate casualties and/or property damages in advance. With the advent of advanced computer technologies, machine learning-based classification models have been employed for managing mountainous disasters, such as landslides and debris flows. This study aims to quantify the soil creep susceptibility using several classifiers, namely the k-Nearest Neighbor (k-NN), Naive Bayes (NB), Random Forest (RF), and Support Vector Machine (SVM) models. To develop the classification models, we downscaled 292 data from 4,618 field survey data. About 70% of the selected data were used for training, with the remaining 30% used for model testing. The developed models have the classification accuracy of 0.727 for k-NN, 0.750 for NB, 0.807 for RF, and 0.750 for SVM against test datasets representing 30% of the total data. Furthermore, we estimated Cohen's Kappa index as 0.534, 0.580, 0.673, and 0.585, with AUC values of 0.872, 0.912, 0.943, and 0.834, respectively. The machine learning-based classifications for soil creep susceptibility were RF, NB, SVM, and k-NN in that order. Our findings indicate that the machine learning classifiers can provide valuable information in establishing and implementing natural disaster management plans in mountainous areas.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.4
• /
• pp.228-234
• /
• 2021

Real estate market participants need to have a sense of predicting real estate prices in decision-making. Commonly used methodologies, such as regression analysis, ARIMA, and VAR, have limitations in predicting the value of an asset, which fluctuates due to unknown variables. Therefore, to mitigate the limitations, an artificial neural was is used to predict the price trend of apartments in Seoul, the hottest real estate market in South Korea. For artificial neural network learning, the learning model is designed with 12 variables, which are divided into macro and micro factors. The study was conducted in three ways: (Ed note: What is the difference between case 1 and 2? Is case 1 micro factors?)CASE1 with macro factors, CASE2 with macro factors, and CASE3 with the combination of both factors. As a result, CASE1 and CASE2 show 87.5% predictive accuracy during the two-year experiment, and CASE3 shows 95.8%. This study defines various factors affecting apartment prices in macro and microscopic terms. The study also proposes an artificial network technique in predicting the price trend of apartments and analyzes its effectiveness. Therefore, it is expected that the recently developed learning technique can be applied to the real estate industry, enabling more efficient decision-making by market participants.