기타 : 1985년도 일본국 학계회고 ; 동양법제사

삼전성만 한일법학회 1986 한일법학 Vol.5 No.-

머신러닝을 적용한 경륜 경기 순위 예측 및 평가에 관한 연구: 2016~2022년 출주표 정보 및 경주 결과 활용

김필수,이상현,전성삼 한국스포츠산업경영학회 2023 한국스포츠산업경영학회지 Vol.28 No.2

본 연구는 기계학습(machine learning)을 적용하여 경륜 경주의 경기 순위를 예측하고, 해당 예측에 활용된 각각의 AI 알고리즘 성능을 비교·분석하기 위하여 실시되었다. 이를 위해 국민체육진흥공단에서 제공한 경륜 선수 출추표와 경기결과 데이터(광명스피돔, 2016~2022년 경주 전수)를 기반으로 실증분석을 시행하였다. 본 연구에서는 파이썬(Python)을 활용하여 자료를 수집하고 가공하였으며, 실질적인 분석을 위해 네이브 베이즈(Naive Bayes), 로지스틱 회귀(Logistic Regression), 랜덤 포레스트(Random Forest), 배깅(Bagging), 서포트 벡터 머신(SVM: Support Vector Machine), 의사결정나무(Decision Tree), 에이다 부스트(AdaBoost), K-최근접이웃(KNN: K-Nearest Neighbor)의 여덟 가지 알고리즘을 적용하였다. 또한, 각 알고리즘의 성능 확인과 평가를 위해 정확도(accuracy), 정밀도(precision), 재현율(recall), F1-score, 평균제곱근오차(RMSE), 결정계수(R-squared)를 평가 기준으로 사용하였다. 본 연구는 단승식, 복승식, 삼복승식을 결과변수로 놓고 실증분석을 시행한 결과 로지스틱 회귀의 성능이 단승식(88.19%)과 복승식(80.07%)의 경우 정확도가 가장 높은 것으로 나타났으며, 삼복승식(78.17%)의 경우 에이다 부스트의 성능이 높게 평가되는 것으로 드러났다. 전체적으로 로지스틱 회귀, 에이다 부스트, 서포트 벡터 머신의 성능이 다른 다섯 가지 알고리즘의 성능과 비교 시 상대적으로 더 우수한 것으로 확인되었다. 본 연구는 경륜 경주의 경주 순위를 예측하기 위해 다양한 머신러닝 알고리즘을 적용하여 이들의 성능을 비교·분석함으로써 기존에는 실증되지 않은 경륜 데이터로 기계학습을 적용하였다는 데 의의가 있다. 본 연구의 실증결과를 통해 스포츠 애널리틱스 분야의 발전을 위해 연구의 방향을 제시하였다는 측면에서의 학문적인 의의는 물론, 실무적인 참고자료로 활용될 수 있을 것으로 판단된다.

Analytical Models for the Initial Stiffness and Plastic Tensile Load Capacity of a Double Angle Connection Subjected to Tension

양재근,전성삼 한국강구조학회 2008 International Journal of Steel Structures Vol.8 No.3

This study has been performed to establish the effects of bolt gage distance and number of bolts on the stiffness and strength of a double angle connection under tension. In addition, analytical models have been proposed to predict the initial stiffness and plastic tensile load capacity of a double angle connection under tension. The effects of prying action have been included in the development of these analytical models. The results obtained from these analytical models have been compared with those obtained from the Thornton model and the Faella model to verify their applications. Six experimental tests have been performed to obtain load-displacement curves of a double angle connection under tension.

영국 프리미어리그 경기데이터 기반 머신러닝을 활용한 경기결과 예측 및 분류모형의 예측 성능 비교

김필수,전성삼,이상현 한국체육학회 2023 한국체육학회지 Vol.62 No.4

The accumulation and collection of big data related to sports along with the development of AI algorithms, and computer science has opened a new era of research on the prediction of sports match results based on machine learning. Football, which has the largest market in the world, calls for such research still in its early stages of academic development and practical needs. This study was conducted to predict the English Premier League match to expand machine learning application in sports result prediction research. To implement our research idea, game match variables were collected from various validated sites such as Premier League, Fotmob, Trasfer market, and Capology homepage through web scraping technique. Following the procedure, 16 variables were selected for the research modelling optimized based on a stepwise selection of a total of 123 variables for 1,107 match data (N=2,214) during the 2020-2021 season to May of the 2022-2023 season collected to predict game results based on nine machine learning algorithms. As a result of the analysis, the prediction performance can be depicted in the order of Multi-layer Perception, Multinomial Logistic Regression, Linear Discriminant Analysis, Quadratic Discriminant Analysis, Support Vector Machine, Gradient Boosting, Ada Boost, Light-GBM, and Random Forest. The Multi-layer Perception, which showed the highest prediction performance, recorded an F1 score of 86.66. This study has made a significant theoretical and practical contribution to the development of game prediction with the application of machine learning algorithms in the domain of professional football, as it has significantly improved the prediction performance of previous studies recording a F1-score of 86.66 that includes expected goal (xG) variable not frequently applied in the prediction of game results in previous studies.

대한민국 축구 국가대표팀 외국인 감독의 위기와 대응: 한국인 스타 코치 선임의약한 유대 효과성

김필수,전성삼,이상현 한국융합과학회 2024 한국융합과학회지 Vol.13 No.3

연구목적 본 연구에서는 대한민국 축구 국가대표팀이 외국인 감독을 선임한 이후 여론이 부정적인상황에서, 이를 해결하기 위해 스타 선수 출신 코치를 코칭스태프로 영입했을 때의 효과를 확인하기위해 웹 기반 빅데이터를 수집하고 분석하였다. 사회적 네트워크 이론에 기반할 때, 약한 유대관계(weak tie)에 있는 사람과의 관계는 친밀감과 정서적 지지를 제공하지는 못하지만, 그 사람이나 집단이지니지 않은 정보와 자원에 접근할 수 있도록 함으로써 개인이나 조직의 성과를 높일 수 있다. 이를검증하기 위해 울리 슈틸리케(Uli Stielike)와 위르겐 클린스만(Jürgen Klinsmann) 감독에 대한 여론이 부정적으로 돌아선 시점에서 이들과 약한 유대관계에 있는 내국인 설기현과 차두리 코치가 코칭스태프로 합류했을 때 위기 국면에서 반등하여 대한민국 축구대표팀의 성과가 개선될 수 있을 것이라상정하였다. 연구방법 본 연구의 실증을 위해 구글(웹, 뉴스, 페이스북), 네이버(블로그, 뉴스, 카페)와다음(블로그, 뉴스, 카페)에서 설기현과 차두리에 대한 네트워크 분석을 시행하고, 이들이 국가대표팀에 코칭스태프로 합류하기 전-후의 여론을 확인하기 위해 워드 클라우드와 감성분석을 실시하였다. 결과 실증분석 결과, 울리 슈틸리케 감독의 지도하에서 스타 선수 출신이었던 차두리와 설기현이 대표팀에 코치로 합류하였을 때 긍정 정서가 각각 37%에서 48%, 39%에서 45%로 증가하였으며, 위르겐클린스만 감독체제 아래에서 차두리가 테크니컬 어드바이저와 코치로 발표된 직후 긍정 정서는 54%에서 63%, 27%에서 38%로 모두 상승한 것을 확인함으로써 외국인 감독이 위기에 직면하였을 때 스타선수 출신 코치의 합류가 여론의 개선에 기여도를 높일 수 있다는 점을 확인할 수 있다. 결론 본 연구는웹 기반 빅데이터를 활용하여 대한민국 축구 국가대표팀에 대한 여론을 샘플로 하여 사회적 네트워크이론을 실증적으로 분석한 연구로 학술 및 실무적 의의를 지닌다. Purpose In this study, we collected and analyzed web-based big data to determine the effectivenessof recruiting and appointing a Korean star coach to the staff to address the negative public opinionduring the tenure of a foreign manager in the Korean national soccer team. Based on social networktheory, relationships with weak ties may not provide intimacy and emotional support, but they canenhance an individual's or organization's performance by providing access to information andresources that the focal person or group does not have. Methods To test this theory, we hypothesizedthat the addition of coaches Seol Ki-hyun and Cha Doo-ri, two Koreans with weak ties to Uli Stielikeand Jürgen Klinsmann, to the coaching staff at a time when public opinion of the coaches had turnednegative, would lead to a rebound from the crisis and improve the performance of the South Koreannational soccer team. We conducted network analysis on Google (web, news, Facebook), Naver (blog,news, cafe), and Daum (blog, news, cafe) for Seol Ki-hyun and Cha Doo-ri, and conducted wordcloud and sentiment analysis to determine public opinion before and after they joined the coachingstaff. Results The empirical analysis shows that positive sentiment increased from 37% to 48% and39% to 45% when Chaduri and Seol Ki-hyun, both former star players, joined the national teamas coaches under Uli Stielike's leadership, and that positive sentiment increased from 54% to 63%and 27% to 38% after Chaduri's announcement as technical advisor and coach under JürgenKlinsmann's leadership, respectively, suggesting that the addition of former star players can contributeto the improvement of public opinion when a foreign coach faces a crisis. Conclusion This studyhas academic and practical implications as an empirical analysis of social network theory usingweb-based big data to sample public opinion on the South Korean national soccer team.

빅데이터 분석을 적용한 대한축구협회와 정몽규 회장의 공정성과 국민 지지 간의 관계에 관한 연구

이상현,전성삼,김필수 한국사회체육학회 2025 한국사회체육학회지 Vol.- No.99

Purpose: This study examines changes in public perception and support for the Korea Football Association (KFA) and its president, Chung Mong-kyu, through big data analysis. The objective of this research is to empirically assess the impact of KFA’s decision-making regarding fairness issues on public support. Method: The research methodology involved collecting text data containing the keyword “Chung Mongkyu” from platforms such as Naver, Daum, and Google. Data were analyzed longitudinally using sentiment analysis, TF-IDF, clustering, and group analysis. Word frequency, TF-IDF, sentiment, and network analyses were conducted on web-based social network data collected from February 2023 to October 2024. Additionally, text analysis was performed on the Korea Football Association and Archery Association using relevant keywords for the same period. Results: The findings of the study indicated that positive perceptions of the Korea Football Association and its president Chung Mong-kyu declined significantly from January 2024 onward. Key factors contributing to the increase in negative perceptions included the appointment of Klinsmann, the withdrawal of the match-fixing amnesty, poor performance at the Asian Cup, Klinsmann’s dismissal, and the subsequent appointment of Hong Myung-bo. Notably, concerns about the fairness of the national team coach selection process substantially influenced the development of negative public opinion. Furthermore, a comparative analysis between the Korea Football Association and the Archery Association revealed that the latter was perceived as having a higher level of fairness during the same period, resulting in a more favorable sentiment towards it. Conclusion: This study empirically confirms that procedural fairness in the decision-making processes of the Korean Football Association significantly influences public support. This finding suggests that maintaining fairness in the management of sports organizations is essential for fostering public trust and support.

Analytical Models for the Initial Stiffness and Plastic Moment Capacity of an Unstiffened Top and Seat Angle Connection under a Shear Load

양재근,전성삼 한국강구조학회 2009 International Journal of Steel Structures Vol.9 No.3

This study was conducted to propose analytical models that can predict the initial stiffness and plastic moment capacity of an unstiffened top and seat angle connection. These analytical models were developed by considering the moment-rotation curves, plastic hinge lines, and failure modes of an unstiffened top and seat angle connection. In the development of these analytical models, the effects of the prying action and the moment-shear interaction of a top angle were also considered. Application feasibility of these analytical models was verified by comparing these models’ results with those obtained from using the Chen-Kishi and Faella models, as well as the Nethercot experimental test data. Four tests were conducted to obtain the moment-rotation curves and failure modes of an unstiffened top and seat angle connection under shear. This study was conducted to propose analytical models that can predict the initial stiffness and plastic moment capacity of an unstiffened top and seat angle connection. These analytical models were developed by considering the moment-rotation curves, plastic hinge lines, and failure modes of an unstiffened top and seat angle connection. In the development of these analytical models, the effects of the prying action and the moment-shear interaction of a top angle were also considered. Application feasibility of these analytical models was verified by comparing these models’ results with those obtained from using the Chen-Kishi and Faella models, as well as the Nethercot experimental test data. Four tests were conducted to obtain the moment-rotation curves and failure modes of an unstiffened top and seat angle connection under shear.

머신러닝 적용 경륜 경주 순위 및 베팅방식별 결과 예측에 관한 연구

김필수,전성삼,이상현 한국서비스경영학회 2023 서비스경영학회지 Vol.24 No.2

The The sports betting industry has proven to be one of the most influential areas in the service sector. Despite its significance, the Korean sports betting industry has been neglected which calls for research scrutiny. This research applies machine learning algorithms (Logistic Regression, Random Forest, AdaBoost, GradientBoost, Light-GBM, Multi-Layer Perceptron, Extra GradientBoost) to predict the results of Keirin competition along with sports betting methods. All of the race data generated in 「Gwangmyeong Speedome」 from 2016 to 2022 were collected and preprocessed for empirical analysis using Python. The results imply that the Logistic Regression had the highest accuracy performance among the machine learning algorithms, with an accuracy of 61.18% for the win prediction, 78.51% for perfecta, 42.37% for the quinella, 31.33% for the exacta, 31.63% for the trio, 22.10% for quinella place, and 14.30% for trifecta bet. Light-GBM and GradientBoost demonstrated the second-highest performance among the machine learning algorithms. In conclusion, this research provides an analysis of the machine learning application of Keirin competition based on sports betting methods. We believe this attempt may contribute to the service management research domain by providing actual prediction results of the sports game to consumers that may to sports betting industry expansion.

현대 프로농구 선수 포지션 역할에 대한 새로운 접근 : 주성분 분석을 통한 NBA(미국프로농구) 선수 포지션 재정의

이상현,전성삼,김필수 한국사회체육학회 2024 한국사회체육학회지 Vol.- No.95

Purpose: This study aims to empirically reclassify the roles of National Basketball Association (NBA) players, moving beyond the traditional categories of point guard (PG), shooting guard (SG), small forward (SF), power forward (PF), and center (C). Method: We utilized the game statistics from three NBA seasons (2020-2023) to redefine players’ positional roles. We employed principal component analysis and regression analysis, utilizing Python and SPSS 26.0 software package for methodological assessment. Results: The empirical results of the principal component analysis showed that modern NBA players can be categorized into six distinct roles: Big Man, First Option, 3-and-Defense, Defensive Guard, Bench Ace, and Role Players. Regression analysis identified 155 players as Role Players, 145 as Big Man, 126 as First Options, 87 as 3-and-Defense players, and 75 as Defensive Guards. Conclusion: This study holds academic and practical importance by systematically collecting, preprocessing, and analyzing extensive NBA data. It redefines player roles based on modern basketball play styles and provides a comprehensive list of players corresponding to these new categories, offering a fresh perspective on player posions.

스포츠 파워랭킹과 머신러닝 기반 선수 역량 평가 및 경기 결과 예측 정확도 고도화

김필수,이상현,전성삼 한국융합과학회 2025 한국융합과학회지 Vol.14 No.8

연구목적 본 연구는 스포츠 경기 외부 환경 변동의 영향을 최소화한 상황에서 선수 개개인의 순수 경기력이나 역량을 정량화하여 객관적으로 등급화 평가할 수 있는 ‘선수 파워랭킹(Power Ranking)’을 체계적으로 개념화하고, 이를 머신러닝 예측모형에 통합함으로써 기존 경기 결과 예측 성능을 뛰어넘는 고도화된 프레임워크를 실증적으로 제시하는 데 그 목적이 있다. 연구방법 본 연구에서는 파이선 3.11.5을 기반으로 2021~2023년 경정 경기데이터(N=13,026)를 수집하고 전처리하였고, 6개의 머신러닝 알고리즘(Light-GBM, 그래디언트 부스트, 다층신경망, 로지스틱 회귀, 랜덤 포레스트, 캣 부스트) 기반의 기본 예측모형을 구축하였다. 이를 바탕으로 경정 선수를 둘러싼 스포츠 경기 외부 환경이 경기 결과에 미치는 영향력을 추산하여 이를 통제한 후, 개별 선수의 파워랭킹을 산출하여 경기 결과 예측모형에 투입하여 성능 개선 효과를 분석하고, 국민체육진흥공단 홈페이지에서 제공하고 있는 AI 예측자료(https://www.kboat.or.kr/rankingpredict)와 예측 정확도를 비교 분석을 시행하였다. 결과 본 연구에서 추산하여 체계화한 선수 파워랭킹 변수를 머신러닝 기반 경정 경기 결과 예측모형에 투입한 결과, 모든 승식에서 예측 정확도가 유의미하게 향상되었다. 이는 현재 국민체육진흥공단 AI 예측자료가 제공하는 경정 예측정보를 상회하는 높은 수준의 예측 성능으로 의미가 있다. 결론 본 연구에서 제시한 선수 파워랭킹은 순수 경기력과 역량을 효과적으로 반영하는 유효한 지표임을 실증적으로 검증하였으며, 스포츠팬 몰입 콘텐츠, 스포츠 베팅 전략 최적화, 스포츠마케팅 등 다양한 방식으로 산업적 활용이 가능하다는 점에서 학술적·실무적 시사점이 존재한다. Purpose This study aims to systematically define a “power ranking” that quantitatively reflects each boat racing athlete’s intrinsic performance and capability under a neutralized environment. By integrating this ranking into machine learning–based prediction models, we seek to empirically demonstrate a framework that outperforms conventional race outcome forecasts. Methods Using Python 3.11.5, we collected and preprocessed 13,026 boat racing records from 2021–2023. We built six baseline prediction models (LightGBM, Gradient Boosting, Multilayer Perceptron, Logistic Regression, Random Forest, and CatBoost). We then estimated the impact of external environmental factors on race results, controlled for these influences, and calculated individual athlete power rankings. These rankings were incorporated into the prediction models to evaluate performance improvements. Finally, we compared our enhanced models’ accuracy with the official AI predictions provided by the Korea Sports Promotion Foundation (https://www.kboat.or.kr/rankingpredict). Results Incorporating the power ranking variable into all win-type prediction models resulted in a statistically significant increase in prediction accuracy. The enhanced models consistently outperformed the Foundation’s existing AI forecast. Conclusion The empirically validated power ranking effectively captures athletes’ pure performance and potential. It holds substantial academic and practical implications, including fan engagement content, optimized betting strategies, and sports marketing applications.