http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
이전제 ( Jeon-je Lee ),조만재 ( Man-jae Cho ),조석헌 ( Seokheon Cho ) 한국정보처리학회 2019 한국정보처리학회 학술대회논문집 Vol.26 No.2
사용자가 보유하고 있는 토큰은 기존 중앙화된 암호 화폐 거래소를 통해 교환 할 수 있다. 그러나 암호 화폐 거래소를 이용한 토큰 교환은 높은 수수료, 암호 화폐 거래소 해킹 가능성, 해당 거래소 내 등록된 암호 화폐에 대해서만 교환이 가능하다는 문제점이 존재한다. 이더리움 플랫폼에 배포된 스마트 컨트랙트는 블록체인 기반으로 다양한 형태의 계약을 조건이 만족할 시 자동으로 이행한다. 본 연구는 스마트 컨트랙트를 이용하여 기존의 중앙화된 암호 화폐 거래소 기반 토큰 교환 문제점을 해결하면서 이더리움 지갑 내 토큰을 사용자가 원하는 ERC-20 토큰으로 자동 교환해 주는 분산형 토큰 교환 시스템을 제안한다.
대규모 언어 모델 (LLM) 기반 COVID-19 백신 접종자별 기저질환 및 알러지 텍스트 데이터 추출 방법에 대한 연구
김솔아(Sola Kim),윤지석(Jiseok Yoon),정다은(Daeun Jeong),김호영(Hoyoung Kim),조석헌(Seokheon Cho) 한국통신학회 2024 한국통신학회 학술대회논문집 Vol.2024 No.6
There has been a significant increase in reported cases of adverse vaccine effects due to the unprecedentedly high vaccination rates resulting from the COVID-19 pandemic. To efficiently manage this vast amount of information concerning not only COVID-19 vaccines but also various other vaccinations, the United States operates and manages the Vaccine Adverse Event Reporting System (VAERS). In this study, we proposed methods for extracting and preprocessing text data on medical history and allergies from the VAERS dataset, which can be used to predict postvaccination adverse effects of COVID-19 vaccines. We structured the text data, grouped individual medical history as well as allergy, and thus created a dataset reflecting individual characteristics by utilizing both large language model (LLM) and various text extraction algorithms. Extracted text data on medical history and allergies can facilitate the understanding of COVID-19 adverse effect of vaccines and serve as key data for effectively responding future adverse reactions.
홍준석(Junseok Hong),이재은(Jaeeun Lee),권민지(Minji Kwon),김동원(Dongwon Kim),조석헌(Seokheon Cho) 한국통신학회 2024 한국통신학회 학술대회논문집 Vol.2024 No.6
This study aims to propose a flight delay prediction model to minimize personal and national economic losses due to unexpected flight delays caused by continuous time-varying weather conditions and to ensure efficient flight schedule operations. We utilized departure flight data from Chicago O’Hare International Airport, known for its high number of flights and departure delay rates, as well as weather data collected at the airport. Due to the imbalance in the dataset used for the flight delay classification prediction model, the downsampling and Synthetic Minority Over-sampling Technique (SMOTE) methods were employed to balance between the majority class of non-delayed flights and the minority class of delayed flights. Additionally, Logistic Regression and Random Forest algorithms were considered for the prediction model. Analysis for our provided aircraft delay prediction models showed that the RF-based model, with SMOTE applied to the dataset including weather data, exhibited the best performance. Furthermore, we introduced the necessity of using the average of the results from training the data separated by different seasons rather than training the entire year’s data without separation by season to improve performance of prediction for the minority class of actual delayed flights.
미국 캘리포니아 교통 데이터를 활용한 인공지능 알고리즘 기반 고속도로 교통량 예측 연구
최석진(Seokjin Choe),김선희(Seonhui Kim),신유민(Youmin Shin),안소현(Sohyeon Ahn),조석헌(Seokheon Cho) 한국통신학회 2024 한국통신학회 학술대회논문집 Vol.2024 No.6
Traffic congestion is one of the major problems that modern cities face. In addition, traffic delays caused by increased traffic flow on highways lead to various adverse effects. In this study, we propose an artificial intelligence algorithmbased model to predict traffic flow using traffic data collected on Highway 78 in San Diego County, California, USA and provided by the California Department of Transportation (Caltrans). Multiple Linear Regression, Random Forest Regression, and Multi-layer Perceptron algorithms were used to predict the traffic flow at a certain location. Moreover, we considered traffic data measured as well as 10-minutes historical data with 30-second or 60-second intervals at its upstream locations to enhance the performance of our proposed prediction models. As a result of our analysis, the traffic flow prediction model based on the Multi-Layer Perceptron algorithm using the historical data with larger intervals showed the best performance.
인공지능 알고리즘을 이용한 은행 예금 가입 의사 분류 예측에 관한 연구
황용우(Yongwoo Hwang),박성환(Sungwhan Park),김재현(Jaehyeon Kim),신현우(Hyeonwoo Shin),조석헌(Seokheon Cho) 한국통신학회 2024 한국통신학회 학술대회논문집 Vol.2024 No.6
This study presents a model based on artificial intelligence algorithms, which predicts deposit subscription intentions using bank customer data in Portuguese. The artificial intelligence algorithms employed for deposit subscription prediction model include Logistic Regression, Random Forest, and Gradient Boost Machine. To resolve data imbalance that is a critical issue in the Portuguese bank user dataset, we utilized various oversampling techniques, such as Synthetic Minority Over-sampling Technique (SMOTE), Borderline SMOTE, and Adaptive Synthetic Sampling for Imbalanced Learning (ADASYN). The objective of our provided models for predicting bank deposit subscription intentions is to accurately identify potential subscribers. A model attains better performance, as it achieves a higher recall that can result in a higher F2 score. Our analysis showed that the Gradient Boost Machine algorithm-based deposit subscription prediction model, employed by ADASYN oversampling, reaches the best performance.
기계학습 알고리즘을 이용한 교통사고 심각도 예측 모델에 관한 연구
박세영(Seyoung Park),송영훈(Younghun Song),김광오(Gwangoh Kim),한결아(GyeolA Han),조석헌(Seokheon Cho) 한국통신학회 2024 한국통신학회 학술대회논문집 Vol.2024 No.6
This study aims to propose a predictive model for the severity of traffic accidents based on external environmental factors, which can perform a role to reduce the number of casualties and accident rates. Data collected from traffic accidents occurring in England from 2021 to 2022 were utilized to provide a traffic accidents severity prediction model. The severity of traffic accidents included in the dataset can lead to a multi-class classification prediction model with the three labels. Moreover, since the severity of most traffic accidents is classified as slight, the dataset exhibits characteristics of imbalanced data. Four artificial intelligence algorithms, such as Adaptive Boosting, Gradient Boosting Tree, K-Nearest Neighbors, and Random Forest, were employed for predicting the severity of traffic accidents. The performance analysis of our prediction models presented that the Random Forest algorithmbased model shows the highest accuracy. However, due to the limitation of imbalanced datasets, the other performance metrics, such as Macro Recall, Macro Precision, and Macro F1-measure, for the Random Forest algorithm-based model showed lower performance compared to accuracy.
이수지(Su-Ji Lee),김규빈(Gyu-Bin Kim),김주현(Joo-Hyun Kim),조석헌(Seokheon Cho) 한국통신학회 2023 한국통신학회 학술대회논문집 Vol.2023 No.6
본 논문은 워크플로 기반 데이터 분석 도구인 KNIME을 활용하여 심장병 분류 예측을 위해 다양한 기계학습 알고리즘들을 적용하여 성능을 비교 및 분석하였다. 이때, Random Forest, XGBoost, Naive Bayes, Multi-Layer Perceptron (MLP) 알고리즘들을 사용하였다. 다차원 이진 데이터인 심장병 데이터 세트에 심장병 분류 예측 성능을 비교 및 분석하기 위해서 4가지의 성능 평가 지표를 고려하였다. 또한, 원본 데이터, Random Forest의 중요도 및 상관관계 속성을 고려하여 선택한 데이터, 그리고 Principal Component Analysis (PCA)를 적용한 데이터 등 3개의 데이터 세트를 생성하여 다양한 분석 결과를 얻었다. 그 결과, 원본 데이터 세트와 속성을 선택한 데이터 세트에서는 MLP 알고리즘이 가장 좋은 성능을 보였다. 또한, PCA를 적용한 데이터 세트에서 XGBoost 알고리즘이 가장 뛰어난 성능을 보였다.