영화 흥행에 영향을 미치는 새로운 변수 개발과이를 이용한 머신러닝 기반의 주간 박스오피스 예측

송정아,최근호,김건우 한국지능정보시스템학회 2018 지능정보연구 Vol.24 No.4

The Korean film industry with significant increase every year exceeded the number of cumulative audiences of 200 million people in 2013 finally. However, starting from 2015 the Korean film industry entered a period of low growth and experienced a negative growth after all in 2016. To overcome such difficulty, stakeholders like production company, distribution company, multiplex have attempted to maximize the market returns using strategies of predicting change of market and of responding to such market change immediately. Since a film is classified as one of experiential products, it is not easy to predict a box office record and the initial number of audiences before the film is released. And also, the number of audiences fluctuates with a variety of factors after the film is released. So, the production company and distribution company try to be guaranteed the number of screens at the opining time of a newly released by multiplex chains. However, the multiplex chains tend to open the screening schedule during only a week and then determine the number of screening of the forthcoming week based on the box office record and the evaluation of audiences. Many previous researches have conducted to deal with the prediction of box office records of films. In the early stage, the researches attempted to identify factors affecting the box office record. And nowadays, many studies have tried to apply various analytic techniques to the factors identified previously in order to improve the accuracy of prediction and to explain the effect of each factor instead of identifying new factors affecting the box office record. However, most of previous researches have limitations in that they used the total number of audiences from the opening to the end as a target variable, and this makes it difficult to predict and respond to the demand of market which changes dynamically. Therefore, the purpose of this study is to predict the weekly number of audiences of a newly released film so that the stakeholder can flexibly and elastically respond to the change of the number of audiences in the film. To that end, we considered the factors used in the previous studies affecting box office and developed new factors not used in previous studies such as the order of opening of movies, dynamics of sales. Along with the comprehensive factors, we used the machine learning method such as Random Forest, Multi Layer Perception, Support Vector Machine, and Naive Bays, to predict the number of cumulative visitors from the first week after a film release to the third week. At the point of the first and the second week, we predicted the cumulative number of visitors of the forthcoming week for a released film. And at the point of the third week, we predict the total number of visitors of the film. In addition, we predicted the total number of cumulative visitors also at the point of the both first week and second week using the same factors. As a result, we found the accuracy of predicting the number of visitors at the forthcoming week was higher than that of predicting the total number of them in all of three weeks, and also the accuracy of the Random Forest was the highest among the machine learning methods we used. This study has implications in that this study 1) considered various factors comprehensively which affect the box office record and merely addressed by other previous researches such as the weekly rating of audiences after release, the weekly rank of the film after release, and the weekly sales share after release, and 2) tried to predict and respond to the demand of market which changes dynamically by suggesting models which predicts the weekly number of audiences of newly released films so that the stakeholders can flexibly and elastically respond to the change of the number of audiences in the film. 2013년 누적인원 2억명을 돌파한 한국의 영화 산업은 매년 괄목할만한 성장을 거듭하여 왔다. 하지만 2015 년을 기점으로 한국의 영화 산업은 저성장 시대로 접어들어, 2016년에는 마이너스 성장을 기록하였다. 영화산업을 이루고 있는 각 이해당사자(제작사, 배급사, 극장주 등)들은 개봉 영화에 대한 시장의 반응을 예측하고 탄력적으로 대응하는 전략을 수립해 시장의 이익을 극대화하려고 한다. 이에 본 연구는 개봉 후 역동적으로 변화하는 관람객 수요 변화에 대한 탄력적인 대응을 할 수 있도록 주차 별 관람객 수를 예측하는데 목적을 두고 있다. 분석을 위해 선행연구에서 사용되었던 요인 뿐 아니라 개봉 후 역동적으로 변화하는 영화의 흥행순위, 매출점유율, 흥행순위 변동 폭 등 선행연구에서 사용되지 않았던 데이터들을 새로운 요인으로 사용하고 Naive Bays, Random Forest, Support Vector Machine, Multi Layer Perception등의 기계학습 기법을 이용하여 개봉 일 후, 개봉1주 후, 개봉 2주 후 시점에는 차주 누적 관람객 수를 예측하고 개봉 3주 후 시점에는 총 관람객 수를 예측하였다. 새롭게 제시한 변수들을 포함한 모델과 포함하지 않은 모델을 구성하여 실험하였고 비교를 위해 매 예측시점마다 동일한 예측 요인을 사용하여 총 관람객 수도 예측해보았다. 분석결과 동일한 시점에 총 관람객 수를예측했을 경우 보다 차주 누적 관람객 수를 예측하는 것이 더 높은 정확도를 보였으며. 새롭게 제시한 변수들을포함한 모델의 정확도가 대부분 높았으며 통계적으로 그 차이가 유의함으로써 정확도에 기여했음을 확인할 수있었다. 기계학습 기법 중에는 Random Forest가 가장 높은 정확도를 보였다.

Forecasting the Equity Risk Premium in the Korean Stock Market: A Factor Analysis Approach

전성주(Sungju Chun) People&Global Business Association 2021 Global Business and Finance Review Vol.26 No.4

Purpose: This article investigates stock return predictability in the Korean stock market using the methodology of dynamic factor analysis. Design/methodology/approach: This article collects monthly data on the equity risk premium on the KOSPI and twelve financial and macroeconomic variables spanning from October 2000 to December 2020 and evaluates the forecasting performance of the dynamic factor predictive regression model by comparing in-sample and out-of-sample predictability with those of individual predictors. Findings: The article finds that the dynamic factor predictive regression exhibits statistically and economically significant in-sample predictability for the future equity risk premium for the KOSPI, as strongly as the best individual predictor can do. Also, the dynamic factor approach can outperform the benchmark historical average in out-of-sample predictability. The detailed analysis of the diffusion indexes reveals that each factor captures different information from various financial and macroeconomic variables relevant for return prediction and the diffusion indexes can deliver better forecasts of the future equity risk premium. Research limitations/implications: There exist different regression methods to combine forecasts comparable to the dynamic factor predictive model such as the forecast combination method by Rapach et al. (2010) and the bagging method by Inoue and Kilian (2008) and Jordan et al. (2017). The study proposes to compare the performance of these models with that of the dynamic factor predictive model in the Korean stock market as future research. Originality/value: The article is the first attempt to apply the dynamic factor predictive regression model to a large set of financial and macroeconomic data in Korea and evaluate its in-sample and out-of-sample predictability in comparison to those of individual predictive variables.

Preoperative Prediction for Early Recurrence Can Be as Accurate as Postoperative Assessment in Single Hepatocellular Carcinoma Patients

차동익,장경미,김성현,김영곤,김한솔,안수현 대한영상의학회 2020 Korean Journal of Radiology Vol.21 No.4

Objective: To evaluate the performance of predicting early recurrence using preoperative factors only in comparison with using both pre-/postoperative factors. Materials and Methods: We retrospectively reviewed 549 patients who had undergone curative resection for single hepatcellular carcinoma (HCC) within Milan criteria. Multivariable analysis was performed to identify pre-/postoperative highrisk factors of early recurrence after hepatic resection for HCC. Two prediction models for early HCC recurrence determined by stepwise variable selection methods based on Akaike information criterion were built, either based on preoperative factors alone or both pre-/postoperative factors. Area under the curve (AUC) for each receiver operating characteristic curve of the two models was calculated, and the two curves were compared for non-inferiority testing. The predictive models of early HCC recurrence were internally validated by bootstrap resampling method. Results: Multivariable analysis on preoperative factors alone identified aspartate aminotransferase/platelet ratio index (OR, 1.632; 95% CI, 1.056–2.522; p = 0.027), tumor size (OR, 1.025; 95% CI, 0.002–1.049; p = 0.031), arterial rim enhancement of the tumor (OR, 2.350; 95% CI, 1.297–4.260; p = 0.005), and presence of nonhypervascular hepatobiliary hypointense nodules (OR, 1.983; 95% CI, 1.049–3.750; p = 0.035) on gadoxetic acid-enhanced magnetic resonance imaging as significant factors. After adding postoperative histopathologic factors, presence of microvascular invasion (OR, 1.868; 95% CI, 1.155– 3.022; p = 0.011) became an additional significant factor, while tumor size became insignificant (p = 0.119). Comparison of the AUCs of the two models showed that the prediction model built on preoperative factors alone was not inferior to that including both pre-/postoperative factors {AUC for preoperative factors only, 0.673 (95% confidence interval [CI], 0.623– 0.723) vs. AUC after adding postoperative factors, 0.691 (95% CI, 0.639–0.744); p = 0.0013}. Bootstrap resampling method showed that both the models were valid. Conclusion: Risk stratification solely based on preoperative imaging and laboratory factors was not inferior to that based on postoperative histopathologic risk factors in predicting early recurrence after curative resection in within Milan criteria single HCC patients.

신경망분석을 이용한 축구경기 승,패 예측모형 개발 -2006독일월드컵 대회를 중심으로-

김주학 ( Joo Hak Kim ),노갑택 ( Gap Taik Ro ),박종성 ( Jong Sung Park ),이원희 ( Won Hi Lee ) 한국스포츠정책과학원(구 한국스포츠개발원) 2007 체육과학연구 Vol.18 No.4

본 연구는 축구 경기 기록자료를 활용하여 축구경기의 승·패를 예측할 수 있는 모형을 개발하는 데 있다. 활용된 경기기록 자료는 2006년 독일 월드컵대회에 나타난 축구경기 전체 64경기의 FIFA TSG공식자료와 M대학교 경기분석 프로그램을 활용하여 산출된 경기기록 자료를 사용하였다. 연구기법으로는 최근에 예측연구에 많이 활용되고 있는 신경망 분석을 이용하여 예측율을 확인 하였으며, 예측율은 2개의 예측모형으로 분류하여 결과를 산출하였다. 구체적인 내용 및 방법은 다음과 같다. 2006년 독일월드컵에 나타난 기록 관련 요인 중 경기 내적요인을 중심으로 축구 경기의 승·패와 관련한 요인을 탐색하고 2006년 독일월드컵 전 경기에 대한 비디오녹화를 실시하였으며, 녹화된 비디오를 통하여 기록요인에 대하여 실시간 축구 경기분석 프로그램을 활용, 승·패와 관련한 기록요인을 점수화하였다. 점수화된 기록요인은 점유율, 성공률 관련 등 여러 가지 요인이 있었으나 실제 경기에서 예측을 할 수 있는 요인을 전문가와 회의를 통하여 선정하고 선정된 요인을 바탕으로 신경망 분석방법을 이용하여 축구경기 승·패 예측 모형을 개발하여 예측율을 확인하였다. 본 연구의 결과는 다음과 같다. 첫째, 전문가회의를 통하여 기록화 할 수 있는 기록자료 요인을 중심으로 2006년 독일월드컵 전체 경기에 대하여 축구 경기분석 프로그램을 통하여 점수화하여 실제 다양한 요인이 내재된 경기에 대하여서도 내용분류를 통하여 점수화할 수 있음을 확인하였다. 둘째, 점수화된 경기 내적요인을 중심으로 신경망분석 방법을 통하여 축구경기에 대하여서도 승·패를 예측할 수 있음을 확인하였으며 타 종목에 대하여서도 실현 가능함을 확인할 수 있었다. 셋째, 축구경기의 승패에 대하여 87.5%의 예측율을 보임으로써 축구경기의 요인에 대한 중요도를 평가할 수 있는 연구의 기초를 확립하였다고 판단된다. This study was on purpose to investigate the various factors of soccer games into inner factors and classify performance factors objectively and scientifically using a soccer game analysis program and develop a model which can predict the result of the soccer games using the Neural NetWork Analysis that stands on the game graded record. The game recorded data was collected from the FIFA TSG official record and the data which was gained using a game analysis program. It has searched the factors related the game results which were focused on the game inner factors of the related factors in 2006 World cup games and recorded the whole games of 2006 World cup games, graded the recording factors related the win/lose using the video and the real time game analysis program with numbers. The graded recording factors with numbers were several; ball possession, completion, and so on. However, through the meeting with Neural NetWork Analysis, professionals prediction related factors which can predict in the real games were selected. And a soccer game win/lose prediction model was developed using the selected factors and Neural NetWork Analysis. It also was compared with the preceding researches in documents record studies. Followings are main results of this study; First, It was confirmed that it is available to grade the real factors in the games which include various factors through the professionals` meetings which implemented the analysis the 2006 Worldcup games and it was focused on the recording factors. Second, the graded recording inner-factors make it possible to predict the soccer game results using the Neural NetWork Analysis method. It was verified that it can be effective in other sports also. Third, the rate of predicting the soccer games` results was 87.5%. It is possible to understand this study contributes on the basis of the importancy of soccer game factors.

AHP 분석을 활용한 수입 자동차 리콜부품의 수요 예측 요인에 관한 연구

정상천(Jung, Sang Chun),김승철(Kim, Seoung Chul),이태원(Lee, Tae Won) 글로벌경영학회 2020 글로벌경영학회지 Vol.17 No.4

본 연구는 국내 자동차 리콜의 부품 수요 예측을 위하여 고려해야 할 요인들에 대해서 그 중요도 분석을 수행하고 있다. 리콜 부품 수요는 리콜의 타입에 따라서 수요 패턴이 달라질 가능성이 높다. 따라서 본 연구에서는 리콜 타입을 분류하는 기준으로서 리콜 관련 요인들의 종류와 그 중요도를 분석하고자 하였다. 자동차 부품 수요 예측을 위한 요인들은 선행 연구 검토와 델파이 조사를 통한 업계 전문가들의 의견을 반영하여 대상 차종의 특성, 리콜 관련 규모, 리콜의 발생 원인, 외부 환경 요인 등 4개의 유형으로 분류하였다. 이러한 요인들의 유형 분류와 도출을 위하여 본 연구에서는 델파이 연구 방법을 수행하였고 추출된 리콜 수요 결정 요인들의 상대적 중요도 분석을 위하여 AHP를 수행하였다. 본 연구에서 수행된 AHP 분석 결과는 다음과 같다. 대분류의 속성 중요도 검증 결과 리콜 발생 원인이 리콜 수요 예측에 가장 중요한 것으로 나타났고, 그 다음은 리콜 관련 규모, 외부 환경 요인, 대상 차종특성 등의 순으로 나타났다. 중분류 속성 항목들에 대한 분석 결과를 살펴보면 리콜 발생 원인중 3개 항목이 1~3위의 순위를 보여주고 있다. 치명성은 전체 중분류 항목들 중에서 가장 중요한 요인 것으로 나타났다. 외부 환경 요인중 정부의 리콜 관련 정책강도를 제외하고는 대상 차량의 판매가격과 함께 전반적으로 상대적 중요도가 모두 낮은 것으로 나타나 리콜 수요를 결정하는 요인으로서 중요도가 낮은 것으로 판단된다. 다음은 딜러사 와 비딜러 두 집단 간 상대적 중요도의 차이를 분석하였다. 분석 결과 딜러사와 비딜러의 경우 모두 전체 결과와 마찬가지로 리콜 발생 원인의 상대적 중요도가 가장 높은 것으로 나타났고 대상 차종 특성이 가장 낮은 것으로 나타났다. 그러나 두 집단에 대한 분석 결과에서는 리콜관련 규모의 가중치가 상이한 것으로 나타났으며 그 이유는 두 집단간 업무 특성에 기인한 것으로 판단된다. 본 연구 결과로서 자동차 리콜 수요를 결정하는 요인들은 리콜의 주요 원인, 리콜 차종의 주력 차종 여부, 리콜 대상 수량, 리콜 유형(자발적 리콜/강제적 리콜), 경쟁사 리콜 여부, 기간별 리콜율, 경기 상황 등과 같은 요인들인 것으로 밝혀졌다. 이러한 요인들은 리콜의 타입 별 부품 수요 예측 모형을 추정하는 데 중요한 정보를 제공하게 될 것이다. 본 연구에서 도출된 리콜 수요 요인들을 토대로 다양한 리콜 타입을 분류하고 각 타입 별 부품 수요의 패턴을 추정할 경우 좀 더 정교한 리콜 부품 수요 예측 모형을 추정할 수 있을 것이다. This study analyzes the priority among factors to consider in predicting the demand for parts for automobile recalls in Korea. The pattern of demand for recall parts depends on the type of recalls. Therefore, this study attempts to analyze the various factors behind recall as the basis of classification of recall types, and their relative priority. Researchers reviewed the prior study and used the Delphi method to seek industry experts’ opinions, and classified the factors for predicting demand for automobile parts into four main categories: properties of target vehicle models, size of recalls, cause of recalls, and external factors. We used the Delphi technique to identify the categories of factors, and applied AHP to analyze the relative priorities of identified factors. The process and the result of the AHP for this study are as follows: among the four main categories, the cause of recalls was found the most important, followed by the size of recalls, external factors, and properties of target vehicle models. Analysis of sub-categories showed the three items under the main category “cause of recalls” ranked the 1st,2nd, and 3rd in terms of importance, on the list of all factors. Fatality is found to be the most critical factor among all sub-category items. Except for the strength of the government’s recall policy, external factors are found less important, along with the sales price of target vehicles, in determining recall demand. Different perception of priorities between dealers and non-dealers was also analyzed. The analysis showed that both dealers and non-dealers find the cause of recalls as the most important, and properties of target models as the least important, which is in line with the overall survey result. However, the two groups apply different weights to the size of recalls, probably because of the difference in what they do. This study concluded that the determinants of recall demand include major reasons for recall, whether the recalled vehicle is a flagship model, the number of recall vehicles, type of recall (voluntary or mandatory), whether competitors are doing a recall, the rate of recall by period, and economic situation. These factors will provide important information on estimating the prediction model for part demand by type of recalls. If recall types are classified and the demand pattern per type is estimated based on the various factors identified through this research, a more accurate model of recall part demand will become possible.

머신러닝 분류 모형을 이용한 FIBA 여자농구 아시안컵 대회의 승패 예측 및 요인 분석에 관한 연구

예원진(Ni, Yuan-Zhen),이성노(Lee, Seong-No) 한국체육과학회 2022 한국체육과학회지 Vol.31 No.5

The purpose of this study is to predict match results and analyze win/loss factors by combining big data and machine learning classification models using the box scores of the 2015-2021 womens basketball Asian Cup tournament. The subject of this study was a total of 200 game records among the records obtained through the official records of the 2015, 2017, 2019, and 2021 Womens Basketball Asian Cup tournaments, and a total of 22 variables were used to predict win/loss results and analyze win/loss factors. In order to predict the win/loss result of the Womens Basketball Asian Cup competition, five machine learning classification models are used, KNN, Decision Tree, SVM, Logistic Regression, and Random Forest, and predictive performance by model by predicting win/loss results. were comparatively analyzed. In addition, in order to analyze the factors affecting win/loss, the importance of each factor was analyzed using a random forest classification model. First, when analyzing factors affecting win/loss using box score data, it was considered that total score and efficiency factors should be removed before analysis in order to obtain more accurate factor importance. Second, in the analysis of factors affecting victory and defeat after cleaning dirty data, the number of successful shots (FGM) was found to be the most important factor, followed by the shot success rate (FG%), the two-point success rate (2PTS%), and personal fouls (PF),interception (STL), and so on. Third, in predicting win-loss results, the logistic regression model showed optimal prediction performance than the KNN, decision tree, SVM, and random forest models, and showed 95% prediction accuracy and 0.95 F1 score.

흉골 골절 환자에서 심혈관계 동반 손상의 예측 인자와 응급 심장 초음파의 적응증

김찬웅,류지영,전영진 梨花女子大學校 醫科大學 醫科學硏究所 2001 EMJ (Ewha medical journal) Vol.24 No.1

Objective : To determine the predicting factors related to cardiovascular injuries and To suggest a clinical indication for emergency echocardiography in sternal fractures. Materials and Results : A total mumber of 40 patients with sternal fractures a over 5-year period were retrospectively assessed on clinical, echocardiographic and biochemical status. We analyzed the following 4 factors as predicting factors for cardiovascular injuries in sternal fractures : 1) presence of restraint, 2) presence of associated injuries, 3) presence of a past medical history involving cardiovascular system, 4) Revised Trauma Score(RTS). We, also, assessed the utility of conventional diagnostic methods for cardiovascular injuries, such as ECG, chest X-ray, and enzyme levels. Based on the methods, we tried to infer an indication for emergency echocardiography in sternal fractures. Results : The presence of a past medical history involving cardiovascular system and abnormal RTS on admission were significant predicting factors. Emergency echocardiography was performed according to the predicting factors and the results from conventional evaluations. These data can suggest that indications for emergency echocardiography in sternal fractures include as 1) if more than two studies reveal abnormality without any significant predicting factors. 2) if more than one study reveal abnormality with any significant predicting factors. Conclusion : The past medical history involving cardiovascular system and initial vital signs imply the presence of associated cardiovascular injuries in sternal fractures. And if possible, emergency echocardiography is recommended.

머신러닝 알고리즘 기반 초등예비교사의 실과 비대면 온라인 수업 만족도 예측 요인 탐색

이철현(Lee, Chul-hyun) 한국실과교육학회 2021 한국실과교육학회지 Vol.34 No.1

이 연구의 목적은 교육대학교 실과 비대면 온라인 수업의 만족도를 조사하고, 머신러닝 알고리즘을 기반으로 만족도에 영향을 미치는 요인의 중요도 및 개별 예측 요인의 의존성을 탐색하는 것이다. 이를 위하여 실과교육 관련 비대면 온라인 강좌를 수강한 교육대학교 1~4학년 학생 331명의 데이터를 수집하여 분석하였다. 수집된 데이터에 대하여 기술 통계를 통해 온라인 수업 만족 예측 요인과 만족도 수준을 분석하였고, 5종의 머신러닝 알고리즘을 적용하여 산출된 머신러닝 모델 성능 및 예측 변수의 중요도를 탐색하였다. 또, 예측 변수에 대한 부분 의존 도표 및 개별 조건부 기대치를 산출하여 학습 만족도에 대한 개별 예측 변수의 의존성을 살펴보았다. 그 결과 K-최근접 이웃과 랜덤 포레스트의 성능이 가장 우수한 것으로 판단되었고, 상호 작용과 학습 동기의 중요도가 상대적으로 높게 나타났다. 또, 예측 변수의 영향이 학습 만족도에 대체로 비례함을 알 수 있었다. 이 연구를 통해 얻은 결론은 다음과 같다. 첫째, 매체 효능감, 교육 지원 서비스, 동영상 강의 품질에 대한 최소한의 만족을 유지하도록 수업의 각 주체가 노력해야 한다. 둘째, 수업의 각 주체는 상호 작용, 자기주도성, 학습 동기의 수준을 지속적으로 향상시킬 수 있도록 주의를 기울여야 한다. 셋째, 교수자는 양질의 실습 환경을 제공하고, 실습 과정의 문제와 질문에 적극적이고 신속하게 대응해야 한다. 넷째, 교수자는 예외적 학습자를 파악하여, 이들에게 적절한 수업 처방을 제공해야 한다. The purpose of this study is to investigate the satisfaction level of non-face-to-face online classes related to Practical Arts Education at the University of Education, and to explore the importance of factors affecting the satisfaction and the dependence of individual predictive factors based on machine learning algorithms. To this end, the data from 331 students in the first to fourth grades of the University of Education who took non-face-to-face online courses related to practical arts education were collected and analyzed. For the collected data, the factors predicting online class satisfaction and learning satisfaction were analyzed through descriptive statistics, and the machine learning model performance and the importance of predictive factors calculated by applying five types of machine learning algorithms were explored. In addition, the dependence of individual predictive factors on learning satisfaction was examined by calculating partial dependence plots and individual conditional expectation for predictive factors. As a result, it was judged that the K-nearest neighbor and the random forest had the best performance, and the importance of interaction and learning motivation was relatively high. In addition, it was found that the influence of the predictive factors was generally proportional to the learning satisfaction. The conclusions obtained through this study are as follows. First, each subject of the class should make efforts to maintain minimum satisfaction with media efficacy, educational support services and the quality of video lectures. Second, each subject in the class should pay attention to continuously improve the level of interaction, self-directedness, and learning motivation. Third, instructors must provide a high-quality practice environment and respond actively and promptly to problems and questions in the practice process. Fourth, instructors must identify exceptional learners and provide them with appropriate instructional practices.

테니스 경기결과 예측 시뮬레이터 설계를 위한 기초연구

최형준(HyongJunChoi),김주학(JooHakKim) 한국체육학회 2009 한국체육학회지 Vol.48 No.4

이 연구는 테니스 경기결과 예측 시뮬레이터 설계를 위한 기초연구로써, 예측 모형의 설계에 있어서 자료의 범위, 자료의 특성, 예측을 할 개인선수의 자료 분산 등을 토대로 예측 적중률의 변화를 살펴보았다. 자료의 범위는 시간적인 배경에 따라서 3개월 이전자료부터 1년 이전자료까지 범주화하였고, 원자료를 이용하여 다중회귀식에 적용한 상황과 Z-점수를 이용하여 다중회귀식에 적용한 사례로 나누어 측정하였다. 2008년 4대 그랜드슬램 테니스 대회(Australian Open, French Open, Wimbledon, US Open)의 경기결과를 공식 사이트를 통해 수집하였고, 2009년 Australian Open의 제 1라운드의 64경기를 예측해보고 실제 경기결과와 비교하였다. 이 연구를 통하여 다음과 같은 결과를 얻을 수 있었다. 첫째, 원자료를 적용하여 다중회귀식을 설계한 경우보다는 상대적인 척도인 Z-점수를 적용하여 다중회귀식을 설계할 경우에 높은 예측 적중률을 나타냈다. 둘째, 예측을 해야 하는 개인선수가 최신 년도에 행한 경기내용이 많을수록 높은 예측 적중률을 나타냈다. 셋째, 다중회귀식의 계산방법에 있어서 전체의 변인을 사용하는 입력방법(Enter method)보다는 예측에 상관성이 높은 변인만을 사용한 후진제거방법(Backward method)이나 전진 제거 방법(Forward method)을 사용할 경우 높은 예측 적중률을 나타냈다. This study was to investigate the foundations for designing predictive simulators based on data range, data characteristics, calculating methods and variances of individual players' performance to prediction. The data range were categorized from 3 month before to 1 year before the prediction. The 2008 Grandslam Tennis tournament were considered for the data collection that it has been used to design the prediction models and comparison of actual match results. Totally, 64 matches in 1st round in 2009 Australian Open were looked up that there were significant results found. As the results of the study, there were several findings. Firstly, to use raw-data for the prediction was not accurate than to use Z-score which is interactive index of performance. Secondly, the amount of individual data involved in the data for prediction was significantly influenced to the prediction results. Thirdly, the reduction of numbers of predictive variables using the Backward and Forward methods were more accurate than to use Raw-data by Enter methods among the multiple regression models.

유전지표를 활용한 사상체질 분류모델

반효정,이시우,진희정,Ban, Hyo-Jeong,Lee, Siwoo,Jin, Hee-Jeong 사상체질의학회 2020 사상체질의학회지 Vol.32 No.2

Objectives Genome-wide association studies(GWAS) is a useful method to identify genetic associations for various phenotypes. The purpose of this study was to develop predictive models for Sasang constitution types using genetic factors. Methods The genotypes of the 1,999 subjects was performed using Axiom Precision Medicine Research Array (PMRA) by Life Technologies. All participants were prescribed Sasang Constitution-specific herbal remedies for the treatment, and showed improvement of original symptoms as confirmed by Korean medicine doctor. The genotypes were imputed by using the IMPUTE program. Association analysis was conducted using a logistic regression model to discover Single Nucleotide Polymorphism (SNP), adjusting for age, sex, and BMI. Results & Conclusions We developed models to predict Korean medicine constitution types using identified genectic factors and sex, age, BMI using Random Forest (RF), Support Vector Machine (SVM), and Neural Network (NN). Each maximum Area Under the Curve (AUC) of Teaeum, Soeum, Soyang is 0.894, 0.868, 0.767, respectively. Each AUC of the models increased by 6~17% more than that of models except for genetic factors. By developing the predictive models, we confirmed usefulness of genetic factors related with types. It demonstrates a mechanism for more accurate prediction through genetic factors related with type.