http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
원종현,신종민,김재호,이장원 한국통신학회 2023 韓國通信學會論文誌 Vol.48 No.6
최근 다양한 분야에서 주목받고 있는 기계학습의 성능은 기계학습 모델의 하이퍼파라미터에 의존한다. 이에 따라 기계학습의 성능을 향상시키기 위해서는 최적의 하이퍼파라미터를 찾는 것이 중요하다. 기계학습의 하이퍼파라미터 최적화는 최적화 문제의 목적 함수와 결정 변수들의 특징 때문에 풀이에 어려움이 많으며, 하이퍼파라미터최적화의 알고리즘들은 이러한 어려움을 해결하는 방향으로 연구되고 있다. 본 논문에서는 기계학습의 하이퍼파라미터 최적화의 난점을 분석하고, 이를 해결하기 위해 제안된 하이퍼파라미터 최적화 연구들의 동향을 파악한다. 또한 이를 바탕으로 기계학습의 성능을 더 향상시키기 위해 앞으로 기계학습의 하이퍼파라미터 최적화 연구가 나아갈 방향을 제시한다.
차세대 칼슘 이온 배터리 양극재 설계 및 선별을 위한 기계학습 플랫폼
김민선(Minseon Kim),박재정(Jaejung Park),김희규(Heekyu Kim),이재준(Jaejun Lee),이인효(Inhyo Lee),이승철(Seungchul Lee),민경민(Kyoungmin Min) 대한기계학회 2023 대한기계학회 춘추학술대회 Vol.2023 No.11
Machine learning was generated for developing highly stable cathode materials with the Ca-Ion Battery NASICON structure. The database is divided into a training set of 146,309 materials and a test set of 630 materials with newly designed NASICON structures. Employing 149 descriptors, including 147 chemical features and 2 structural features derived from the composition of each material. Random forest (RF) regressor, employed for Eform prediction, demonstrated impressive results with an R-squared of 0.916, MAE of 0.142, and RMSE of 0.351 eV/atom. Similarly, the RF classifier used for Ehull prediction exhibited an Accuracy of 0.818, AUC of 0.889, and Precision of 0.826. The optimal model was subsequently applied to predict stable materials among the 630 materials, based on the criteria of (1) Eform < 0 eV/atom and (2) Ehull < 0.05 eV/atom. As a result, 125 materials were identified as possessing both structural and thermodynamic stability in charge and discharge states.
윤명섭(Myung-Sup Yoon),이동혁(Dong-Hyuk Yi),윤원식(Won-Sik Yoon),서명교(Myung-Kyo Seo),유승엽(Seung-Yup Ryu) 대한기계학회 2022 대한기계학회 춘추학술대회 Vol.2022 No.11
Supervised machine learning technique was applied to accurately predict the performance of the clean room air conditioner (CRAC) installed in the field. The performance of two neural networks was compared. One is the control group neural network using the laboratory sensor data and the other is the experimental group neural network using the product sensor data as an input. In both cases, they share laboratory performance results as an output label. Training data set of 2,816 combinations were acquired in the laboratory for the various indoor climate, outdoor climate and CRAC fan output conditions. When predicted with two trained ANNs, the control group showed better results thant the experimental group. In addition, the experimental group ANN performance prediction showed relatively more accurate results than the performance values calculated directly from the product sensors.
윤명섭(Myung-Sup Yoon),이동혁(Dong-Hyuk Yi),윤원식(Won-Sik Yoon),서명교(Myung-Kyo Seo),유승엽(Seung-Yup Ryu) 대한기계학회 2022 대한기계학회 춘추학술대회 Vol.2022 No.11
Supervised machine learning technique was applied to accurately predict the performance of the clean room air conditioner (CRAC) installed in the field. The performance of two neural networks was compared. One is the control group neural network using the laboratory sensor data and the other is the experimental group neural network using the product sensor data as an input. In both cases, they share laboratory performance results as an output label. Training data set of 2,816 combinations were acquired in the laboratory for the various indoor climate, outdoor climate and CRAC fan output conditions. When predicted with two trained ANNs, the control group showed better results thant the experimental group. In addition, the experimental group ANN performance prediction showed relatively more accurate results than the performance values calculated directly from the product sensors.
머신러닝(XGBoost)기반 미국프로야구(MLB)의 투구별 안타 및 홈런 예측 모델 개발
조선미,김주학,강지연,김상균 한국체육측정평가학회 2023 한국체육측정평가학회지 Vol.25 No.1
Recently, research on artificial intelligence-based prediction of baseball has been gradually developed, and various studies are being conducted. In general, variables such as launch angle and launch speed are used to predict hitters' home-run and hit. However, launch angle and launch speed(batting results) are correlated with the pitcher's pitching. Therefore, in this study, the artificial intelligence model that predict home-run and hit was developed using only the pitcher's pitching information, excluding the batting information of the hitter. To develop the artificial intelligence model, pitching data from the 2022 season of the Major League Baseball(MLB) collected. (independent variable : pitch type, release speed, zone, stand, pitch throws, ball count, strike count / dependent(prediction) variable : hit and home-run) The artificial intelligence model was developed using XGBoost, one of the machine learning method. As a result of the development, the artificial intelligence model performance was accuracy 86.92%, precision 99.90%, recall 86.95%, and F1 score 92.98%. Ryu Hyun-jin(MLB pitcher)'s pitching data was applied to the developed artificial intelligence model and confirmed characteristics of individual. 최근 야구 종목의 인공지능 기반의 예측 연구는 점진적인 발전을 보이며, 다각도의 연구가 진행되고 있다. 보편적으로 타자의 홈런과 안타를 예측하기 위해서 타구의 발사각도, 타구속도 등의 요인이 활용된다. 그러나 타구결과인 발사각도 및 타구속도는 투수의 투구와 상관이 높다. 따라서 이 연구에서는 타자의 타격정보를 제외하고, 투수의 투구정보만을 활용하여 홈런과 안타를 예측하는 인공지능 모델을 개발하였다. 모델개발을 위해 2022시즌 미국프로 야구(MLB)의 투구데이터를 수집하였으며, 구종, 투구존, 타자위치, 투수주손, 볼카운트, 스트라이크카운트의 변수를 독립변수로 홈런과 안타 결과를 종속변수(예측변수)로 하는 인공지능 모델을 개발하였다. 인공지능 모델은 머신러닝 기법 중 하나인 XGBoost를 사용하여 개발하였다. 개발 결과, 정확도 86.92%, 정밀도 99.90%, 재현율 86.95%, F1 score 92.98% 성능의 인공지능 예측 모델이 개발되었다. 아울러 개발 모델에 미국프로야구(MLB)의 투수인 류현진의 투구데이터를 적용하여 데이터 탐색적 분석을 수행하여, 개인 선수의 보편성과 특징을 확인하였다.
조영창,이홍재,최용락 한국IT정책경영학회 2017 한국IT정책경영학회 논문지 Vol.9 No.6
Artificial Neural Network studies mimicking the current human cortex, Visual Cortex or Auditory Cortex, are underway. However, there is a lack of empirical research on the union domain that integrates, memorizes, and judges the information from each sensory domain. So, in this paper, we want to study 'Value-category Memory' related to the coalition domain. The value-category memory term comes from Gerald Edelman's brain higher-order consciousness model and is a system that categorizes and memorizes according to the value of the object. In 'Value-category Memory', value utilizes cognitive semantics, a linguistic tool related to cognition. In this paper, we try to design a 'Value-category Memory' system that interprets the meaning of language through physical experience.
Forecasting the Busan Container Volume Using XGBoost Approach based on Machine Learning Model
웬티프엉타인,조규성 한국사물인터넷학회 2024 한국사물인터넷학회 논문지 Vol.10 No.1
Container volume is a very important factor in accurate evaluation of port performance, and accurate prediction of effective port development and operation strategies is essential. However, it is difficult to improve the accuracy of container volume prediction due to rapid changes in the marine industry. To solve this problem, it is necessary to analyze the impact on port performance using the Internet of Things (IoT) and apply it to improve the competitiveness and efficiency of Busan Port. Therefore, this study aims to develop a prediction model for predicting the future container volume of Busan Port, and through this, focuses on improving port productivity and making improved decision-making by port management agencies. In order to predict port container volume, this study introduced the Extreme Gradient Boosting (XGBoost) technique of a machine learning model. XGBoost stands out of its higher accuracy, faster learning and prediction than other algorithms, preventing overfitting, along with providing Feature Importance. Especially, XGBoost can be used directly for regression predictive modelling, which helps improve the accuracy of the volume prediction model presented in previous studies. Through this, this study can accurately and reliably predict container volume by the proposed method with a 4.3% MAPE (Mean absolute percentage error) value, highlighting its high forecasting accuracy. It is believed that the accuracy of Busan container volume can be increased through the methodology presented in this study.