Earticle

Cardio-Vascular Diseases (CVD) is found to be rampant in the populace leading to fatal death. The statistics of a recent survey reports that the mortality rate is expanding due to obesity, cholesterol, high blood pressure and usage of tobacco among the people. The severity of the disease is piling up due to the above factors. Studying about the variations of these factors and their impact on CVD is the demand of the hour. This necessitates the usage of modern techniques to identify the disease at its outset and to aid a markdown in the mortality rate. Artificial Intelligence and Data Mining domains have a research scope with their enormous techniques that would aassist in the prediction of the CVD priory and identify their behavioural patterns in the large volume of data. The results of these predictions will help the clinicians in decision making and early diagnosis, which would reduce the risk of patients becoming fatal. This paper compares and reports the various Classification, Data Mining, Machine Learning, Deep Learning models that are used for prediction of the Cardio-Vascular diseases. The survey is organized as threefold: Classification and Data Mining Techniques for CVD, Machine Learning Models for CVD and Deep Learning Models for CVD prediction. The performance metrics used for reporting the accuracy, the dataset used for prediction and classification, and the tools used for each category of these techniques are also compiled and reported in this survey.

One of the most common task of electromyography (EMG)-based human?machine interface (HMI) is the hand gesture recognition. Robust and accurate predictions leveraging the surface EMG (sEMG) is a key challenge. In this paper, we present a 3D hand gesture prediction application, leveraging the sEMG signal and the optical hand tracking information. A transformer?encoder classifier (TEC) module is introduced in an IPC-system to predict the 3D-hand gestures using eight monopolar channels from sEMG as input. An experimental testbed is setup to acquire, train, and predict the 3D hand gestures within a feasible range of performance. The performance has been evaluated in terms of percentage of correctly classified keypoints (PCK). PCK is measured by first estimating the euclidean distance between the actual and the predicted keypoints. The percentage of keypoints within a threshold distance value are then calculated. Results from the ablation study indicate that the proposed scheme shows a percentage of correctly classified keypoints of up to 72.8%, 92.7%, 97.2%, and 98.6% with a PCK threshold of 5 mm, 10 mm, 20 mm, and 30 mm, respectively.

Predicting disaster classification on time is critical to mitigate the damage. Since identifying disaster types requires large amounts of data, and real-world data are often imbalanced, there are many recent works addressing data imbalance problems using generative models. However, if the process of generating text data based on disaster classes and severity is not handled improperly, the quality of the data can be degraded as well as the performance of classification predictions. In this paper, we propose a scheme for generating data with enhanced quality using text based on labels such as informational value of text and severity of disasters. Our experiment results verify the quality of data through the comparisons of prediction performance between various machine learning models.

In this paper, we propose the compared performance of normalization methods-based machine learning classification some techniques for NG leak prediction. The natural gas (NG), mostly methane leaks into the air, it is a big problem for the climate. The proposed method is OrdinalEncoder(OE) based K-means clustering and OE transformation based SVM and MLP classifications for predicting NG leak. We have shown that our proposed OE based SVM method accuracy 97.82%, F1-score 98.54% and both of two normalization based MLP accuracy and F1-score also more than 96% which is relatively higher than the other methods. The system has implemented SPSS and Python, including its performance, is tested on real open data.

In horse management, the alarm system with sensors in the foaling period enables the breeder can appropriately prepare the time of the parturition. It is important to prevent losses by unpredictable parturition because there are several high risks such as dystocia and the death of foals and mares during foaling. However, unlike analysis in the alarm system that detects specific motions has been widely performed, analysis of classification following specific behavior patterns or number needs to be more organized. Thus, the objective of this study is to classify signs of the specific behaviors of the mares for the prediction of pre-foaling behaviors. Five Thoroughbred mares (9-20 yrs) were randomly selected for observation of the prefoaling behaviors. The behaviors were monitored for 90 min that was divided into three different periods as 1) from -90 to -60 min, 2) from -60 to -30 min, 3) from -30 min to the time for the discharge of the amniotic fluid, respectively. The behaviors were divided into two different categories as state and frequent behaviors and each specific behavioral pattern for classification was individually described. In the state behaviors, the number of mares in the standing of the foaling group (3.17 ± 0.18b) at period 3 was significantly higher than the control group (1.67 ± 0.46a). In contrast, the number of the mares in the eating of the foaling group (1.17 ± 0.34b) at period 3 was significantly lower than the control group (3.33 ± 0.46a). In the frequent behaviors, the weaving of the foaling group was significantly higher than the control group, and looking at the belly of the foaling group was significantly lower than the control group. In period 2, defecation, weaving, and lowering the head of the foaling group were significantly higher than the control group, respectively. In period 3, sitting down and standing up, pawing, weaving, and lowering the head in the foaling group were also significantly higher than the control group. In conclusion, the behavior is significantly different in foaling periods, and the prediction of foaling may be feasible by the detection of the pre-foaling behaviors in the mares.

Accuracy improvement of classification model becomes main research objective in various fields. Selecting important features and removing outliers of a dataset are two effective solutions for improving model accuracy. Information Gain is one of the feature selection methods that can be considered as a solution for selecting important features of a dataset. Information Gain selects the variable that maximizes the information gain, which in turn minimizes the entropy and best splits the dataset into groups for effective classification. Aside of selecting important feature, removing outlier is also necessary for improving accuracy of the classification model. Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is one of the powerful outlier removal methods which can identify with significant accuracy the clusters of random shape and size in large databases corrupted with noise. Therefore, in this study, we propose the accuracy improvement of heart disease classification model using Information Gain and DBSCAN applied to various machine learning algorithms. One publicly available heart disease dataset (Cleveland) is utilized in this study to build the classification model. The results showed that after implementing Information Gain, the accuracy of the model applied to Gaussian Naïve Bayes, Logistic Regression, Multi-Layer Perceptron, Support Vector Machine, Decision Tree, Random Forest, and Extreme Gradient Boosting algorithms increases as much as 1.31% in average. The accuracy also increases when DBSCAN is applied to the model after utilizing Information Gain, with the number of improvements is around 0.62%.

본 연구는 소득구간별 아동ㆍ청소년의 사회관계에 의한 건강행위 분류예측에 관한 변수 중요도를 분석하기 위해 데이터마이닝 기법을 활용하였다. jamovi 및 R(ver.4.1.0)/R-studio 분석도구로 활용하여 한국아동ㆍ청소년패널조사2018 패널자료의 2차 년도(2019년) 데이터 자료(아동ㆍ청소년 총5,197명과 그들의 보호자 5,197명)를 분석하였고 연구결과는 다음과 같다. 첫째, 혼합분포군집분석을 통해 아동ㆍ청소년의 건강행위 특성에 의해 불건강군집(1군집)과 건강군집(2군집)으로 분류되었으며, 건강군집이 불건강 군집에 비해 주관적 건강인식과 신체활동정도, 이침 식사 정도가 높게 나타났다. 둘째, t 검정 결과 교사관계, 부모양육태도, 부모와 함께하는 시간, 어머니 학력, 부모의 아침식사여부, 가구소득에서 유의미한 차이가 나타났다. 셋째, 소득구간별 아동ㆍ청소년의 건강행위 특성에 영향을 미치는 사회관계요소가 소득상위구간에 유리하게 작용됨을 확인할 수 있었다. 그리고 사회관계로써 친구관계, 교사관계, 부모관계 그리고 부모요인(학력, 신체활동, 아침식사여부) 중에서 부모의 아침식사여부가 모든 소득구간에서 가장 중요한 변수로 나타났다.

In this study, data mining technique was used to analyze the importance of variables in predicting the classification of health behaviors by social relations of children and youth by income category. Using jamovi and R(ver.4.1.0)/R-studio as an analysis tool, the second year (2019) data of the Korean Children and Youth Panel Survey 2018 panel data (a total of 5,197 children and youth and their guardians 5,197) was analyzed, and the results of the study are as follows. First, through mixed distribution cluster analysis, children and youth were classified into unhealthy clusters (cluster 1) and healthy clusters (cluster 2) according to the health behavior characteristics of children and youth. Eating was found to be high. Second, as a result of the t test, significant differences were found in teacher relationship, parenting attitude, time spent with parents, mother’s educational background, whether parents had breakfast, and household income. Third, it was confirmed that the social relation factors that affect the health behavior characteristics of children and youth by income bracket acted favorably in the upper income bracket. And among the social relationships, friend relationship, teacher relationship, parent relationship, and parental factors (educational background, physical activity, breakfast status), whether parents had breakfast was the most important variable in all income categories.

본 연구는 고속도로에서 발생한 교통사고 블랙박스 영상을 기반으로 군집분석과 예측모형 비교를 수행하였다. 분석자료로 사고 직전의 도로 및 교통 상황을 파악할 수 있는 차량 주행행 태, 노면 상태 등 사고 영상에서 추출이 가능한 항목을 설명변수로 활용하였다. 여러 요소에 의해 영향을 받는 교통사고 데이터의 특징을 고려하여 데이터의 이질성을 반영하는 군집분석 을 활용하였다. 군집분석으로 분류된 각 군집을 사고 심각도 수준의 비율을 기준으로 나누고, 종속변수인 인명피해 수준을 반영하여 사고 예측 평가를 수행하였다. 사고 예측모형은 로짓 모형(Logit model)을 적용한 결과, 전체 데이터를 분석한 경우보다 군집분석에 의해 두 개의 사고 심각도 그룹을 분류하여 예측했을 때 우수한 예측 능력을 보여주었다. 이는 군집분석을 통한 그룹별 사고 특성과 사고 심각도를 반영하여 사고위험을 예측하는 것이 더 효과적인 것 으로 판단된다. 또한 2차 사고와 같은 정차 중 추돌사고, 차로변경 중 측면 추돌사고 등이 중요 한 주행행태변수로 작용하는 것으로 나타났다.

This study was based on the black box images of traffic accidents on highways, cluster analysis and prediction model comparisons were carried out. As analysis data, vehicle driving behavior and road surface conditions that can grasp road and traffic conditions just before the accident were used as explanatory variables. Considering that traffic accident data is affected by many factors, cluster analysis reflecting data heterogeneity is used. Each cluster classified by cluster analysis was divided based on the ratio of the severity level of the accident, and then an accident prediction evaluation was performed. As a result of applying the Logit model, the accident prediction model showed excellent predictive ability when classifying groups by cluster analysis and predicting them rather than analyzing the entire data. It is judged that it is more effective to predict accidents by reflecting the characteristics of accidents by group and the severity of accidents. In addition, it was found that a collision accident during stopping such as a secondary accident and a side collision accident during lane change act as important driving behavior variables.

오류의 원인을 이해하고 예측하는 것은 일상생활이나 산업현장에서 일어날 수 있는 사고를 방지하는 데 중요하다. 최근 오류 예측 연구에서는 고정된 시행 간 간격(inter-trial interval, ITI)의 인지 과제를 사용하여 성공적으로 오류를 예측하였지만, ITI을 고려하지 않아 오류 예측 모형을 일반적인 상황까지 적용하기가 어려움이 있다. 본 연구에서는 오류 이전 여섯 시행들로부터 얻어진 반응시간과 ITI 추세, 그리고 오류 시행에서의 과제조건을 이용하여 기계학습을 통해 오류를 예측하고자 하였다. 그 결과, 다양한 오류의 유형이 분류 및 예측되었으며, 특히 ITI 추세에서 반복적인 ITI 패턴이나 시간 압력의 변화와 같은 요인들이 오류와 관련되는 것으로 나타났다. 본 연구는 변화하는 ITI 변인이 오류예측에 중요한 역할을 할 뿐만 아니라 오류 발생과 관련된 피험자의 정신적 상태와 관련이 있다는 것을 확인하였다. 따라서 본 연구는 다양한 환경에서 다양한 유형의 오류를 예측할 수 있으며, 이러한 오류들이 다양한 원인으로 발생할 수 있음을 확인하였다는데 의의가 있다.

It is important to understand causes of human errors and to predict errors in order to prevent various accidents in our daily lives and industrial fields. Although a previous study employing a fixed inter-trial interval (ITI) in a cognitive task successfully predicted errors, it is unlikely to generalize from the previous results to other situations. The current study sought to predict errors by reaction times, task conditions, and ITIs extracted from six consecutive trials preceding error trials, in the context of machine learning. The results showed that various types of errors could be observed and predicted. Especially, presence of repeated patterns or time-pressure chances in the ITI trends might be related to errors. This is interpreted that ITI variation is important to predict errors as well as related to participants’ mental states affecting errors. Therefore, this study suggests that various types of errors in a variety of situations can be predicted, in which those errors would be caused by various factors.

목적 : 본 연구의 목적은 Coronavirys Disease(COVID-19) 팬데믹을 거친 고령자의 사회적 건강의 수준의 변화궤적에 대한 잠재계층을 분류하고 잠재 계층별 특성을 분석함에 있다. 또한 이러한 사회적 건강에 영향을 미치는 예측요인을 파악하여 고령자의 사회적 건강을 증진을 위한 기초자료를 마련하고자 한다. 연구방법 : 국내 고령자의 사회적 친밀도에 따른 사회적 건강 유형을 파악하기 위해서 한국복지패널의 3 년차 종단자료를 토대로 분석하였으며, 연구대상자는 세 시점 모두를 응답한 2845명의 고령자를 대상으 로 하였다. 대상자중심접근인 성장혼합모형(Growth Mixture Model; GMM)을 적용하여 변화궤적에 따 른 잠재계층을 분석 하였고, 도출된 각 잠재유형별 특성을 파악하기 위해 χ2 분석, 분산분석을 실시하 였으며, 계층 간 차이를 유발하는 요인을 파악하기 위해 다항로지스틱 회귀분석을 실시하였다. 결과 : GMM 적용결과, 사회적 건강의 변화궤적에 대한 잠재계층은 최종 4개의 집단으로 저수준 감소-증 가 집단, 중수준 유지-증가 집단, 고수준-감소 집단, 고수준 유지’집단으로 분류되었다. 또한 사회적 건강 수준에 따라 여가만족도에서 차이가 나타나는 것으로 드러났으며, 그 외에도 연령 차이가 존재하였 다. 잠재계층분류에 영향을 미치는 영향변인을 검증한 결과, 특히 여성일수록, 종교를 가지고 있을수록, 여가만족도와 전반적 만족도가 모두 높을수록 고수준 유지 집단에 속할 확률이 높은 것으로 나타났다. 결론 : 국내 고령자의 사회적 건강은 시간이 지남에 따라 감소하는 궤적을 보이는 것으로 나타났다. 변화 궤적에 따라 4개의 집단으로 구분 지을 수 있으며, 각 잠재 유형별 연령과 여가 만족도 부분에서 집단별 차이가 드러났다.

Objective: The purpose of this study is to identify the latent classes for a change in trajectory of the level of social health of older adults who have gone through the COVID-19 pandemic and analyze the characteristics of each class, thereby identifying predictive factors affecting the social health of older adults. Methods: To analyze the types of changes in the trajectory of the level of social health, Korean Welfare Panel Study (KOWEPS) data (2019~2021) was used. The subjects of the study were 2845 older adults who responded at all three time points. The Growth Mixed Model (GMM), which is a subject-centered approach, was applied to analyze the latent classes. In addition, χ2 and ANOVA were conducted to identify the characteristics of each latent type, and a multinominal logistic regression was applied to analyze the influencing factors between classes. Results: As the results of applying GMM indicate, the latent classes for the changes in trajectory of social health were classified into four groups: a low level decrease-increase group, a medium level maintenance-increase group, a high level decrease group, and a high level maintenance group. There was a difference in leisure satisfaction and demographic between classes. In terms of the influencing variables, it was analyzed that the greater the number of women, the more religious the individuals, and the higher their leisure and overall life satisfaction, the higher their probability of belonging to a high level maintenance group. Conclusion: By applying GMM, this study identified the latent class of the changes in trajectory of the social health level of older adults, and it was found that there were significant differences in leisure satisfaction according to the various types. In addition, the influencing factors that distinguish social health were also analyzed.

본 연구는 유전 알고리듬에 기반한 새로운 도산예측기법을 개발하고 그 기법의 타당성 및 예측 우수성을 검증하는데 목적이 있다. 본 연구에서 제안하는 이진분류기법은 도산기업과 비도산기업을 대표할 수 있는 가상기업(virtual company)을 설정하고, 그 가상기업과 분류대상 기업 간의 유사도를 측정하여 도산여부를 분류하는 방법론으로, 가상기업의 변수 값과 각 변수의 가중치는 훈련용 자료의 분류정확도를 극대화할 수 있도록 유전 알고리듬을 이용하여 구하게 된다. 본 연구에서 제안하는 기법의 타당성을 검증하기 위해 기존의 도산예측기법과 예측성과를 실험을 통해 비교한 결과, 본 연구에서 개발한 기법의 예측력이 기존의 다변량판별분석, 로지스틱 회귀모형, 의사결정나무, 인공신경망 모형보다 높은 수준을 보이는 것을 확인하였다.

본 연구의 목적은 노인의 주거빈곤 유형화를 통해 다양하고 복합적인 노인 주거빈곤 특성을 보다 면밀하게 이해할 필요성을 제시하고, 각 유형별 특성 및 예측요인을 분석하는 것이다. 분석자료는 한국복지패널(KOWEPS) 14차년도 자료(2019년)이며, 만 65세 이상 노인 가구주 3,141명을 분석대상으로하였다. 선행연구 검토를 통해 대표적인 주거빈곤 지표와 영향요인을 선정하였고, 분석방법으로 잠재계층분석, 다항로지스틱 회귀분석을 실시하였다. 분석결과 노인의 주거빈곤은 ‘주거점유불안정형(20.7%)’, ‘설비 및 구조 미충족형(3.5%)’, ‘주거안정형(75.7%)’의 3개의 유형으로 확인되었다. 주거빈곤 유형별 인구학적, 사회경제학적, 주거 특성에 따라 차이를 보였다. 공적연금미수급은 주거빈곤 가능성을 증가시키는 공통요인으로 나타났고, 비지상거주는 ‘주거점유불안정형’에, 단독주택거주는 ‘설비 및 구조 미충족형’에 속할 위험을 높이는 주요 예측요인으로 확인되어 유형별 주거빈곤을 예측하는중요 지표로 사용될 수 있음을 보여준다. 연구결과를 토대로 노인의 주거빈곤 위험에 대한 정책적 우선순위 설정, 비지상거주 노인의 주거안정 및 단독주택거주 노인의 주거환경 개선을 위한 제도적⋅법적 장치마련, 노후소득보장정책 개선 등 노인의 주거빈곤 특성을 고려한 주거지원정책의 필요성을 제언하였다.

This study aims to suggest the need to better understand the diverse and complex characteristics of housing poverty among the elderly through their classification types. Additionally, the study establishes principal data for the preparation of housing poverty policy for the elderly by analyzing its predictive factors by type. Data from the 2019 Korea Welfare Panel Study Year 14 were used for analysis, and the subjects were 3,141 household owners aged 65 or older. Representative housing poverty indicators and influencing factors were selected through a review of previous studies, and latent class analysis and multinomial logistic regression analysis were performed as analysis methods. The results showed that housing poverty among the elderly was classified into three groups: “housing occupancy anxiety type (20.7%),” “facility and structure unfilled type (3.5%),” and “housing stability type (75.7%).” There were differences according to the demographic, socioeconomic, and residential characteristics of each type of housing poverty. The non-receipt of public pensions is a common factor that increases the possibility of housing poverty. Non-ground-floor residences have a strong effect that increases the risk of falling into the “housing occupancy anxiety type” and detached house dwellings into the “facility and structure unfilled type,” indicating that they can be used as important indicators to predict housing poverty by type. The results of this study suggest the need for a housing welfare policy considering the housing poverty characteristics of the elderly, such as setting appropriate policy priorities, providing institutional and legal mechanisms for stabilizing the housing of the elderly living on the ground, and improving the residential environment of the elderly living in single-family homes, improving the old age income security policy, etc.

본 연구는 고교야구 투수의 경기 기록 데이터를 활용하여 프로야구 지명 여부를 예측하는 머신러닝 기반 분류 모델 을 개발하는 것을 목표로 한다. 이를 위하여 최근 생성 모델링 분야에서 주목받는 Diffusion 모델의 원리를 도입하 고, 학습 데이터에 점진적으로 노이즈를 주입한 후 다층 퍼셉트론(MLP) 기반 네트워크를 통해 이를 제거하는 과정 을 학습하는 조건부 노이즈 예측 모델을 설계하였다. 본 모델은 소프트맥스(Softmax) 함수를 출력층에 적용한 다 중 클래스 분류 구조로, 고교 투수의 프로 지명 가능성을 '유력', '가능', '어려움'의 세 가지 범주로 예측한다. 모델 입력에는 평균자책점(ERA), 피안타율(BAA), 9이닝당 탈삼진 수(K/9)를 포함한 주요 투수 지표뿐만 아니라, 구 속, 이닝 소화율, 경기 출장 수 등 다양한 정량적 통계 지표가 활용된다. 나아가, Diffusion 모델의 구조적 특성을 통해 예측 과정에서의 불확실성을 정량화하고, 이를 기반으로 보다 신뢰성 있는 예측 결과를 도출할 수 있는 가능성 을 탐색하였다. 제안된 모델은 향후 프로야구 스카우팅 및 선수 평가 과정에서 객관적인 의사결정을 지원하는 도구로 활용될 수 있다.

This study aims to develop a predictive model to determine the likelihood of high school baseball pitchers being drafted into professional leagues based on their performance records. To achieve this, we propose a multi-class classification model that leverages the principle of diffusion modeling, wherein noise is gradually added to the training data during training, and a multi-layer perceptron (MLP) network is trained to denoise the input. The model incorporates a Softmax activation function at the output layer to classify pitchers into one of three categories: highly likely, possible, or unlikely to be drafted. The input features used in the model include core pitching statistics such as Earned Run Average (ERA), Batting Average Against (BAA), and Strikeouts per Nine Innings (K/9), as well as other quantitative indicators like pitching velocity, number of appearances, and innings pitched ratio. Furthermore, the model explores the potential of diffusion-based approaches to quantify and calibrate prediction uncertainty, offering a new perspective on how generative modeling techniques can enhance the reliability of classification outcomes. The proposed model can be utilized as a tool to support objective decision-making in future professional baseball scouting and player evaluation processes.

낙상사고는 많은 재활 비용을 필요로 하고, 다른 안전사고에 비해 대부분 예방이 가능하다는 점에서 올 바른 정보제공을 통한 사전 예방이 더욱 중요하다. 특히, 고령자의 경우 연령증가에 따른 신체의 노화 및 질병은 안전사고의 증가로 이어질 수 있고 상해의 정도도 심각한 실정이다. 따라서, 본 논문에서는 환자에 부착된 통합 센서 모듈에서 실시간으로 수집되는 데이터를 기반으로 낙상 사고에 대한 예측 데이터 모델링을 통해 실시간으로 낙상에 대한 모니터링이 가능한 데이터 분석을 연구하였다. 이를 위해 온도 센서, 심박 센서 및 모션 센서에서 실 시간으로 데이터를 수집하고 이항 분류 모델링을 적용한 낙상 예측 데이터 분석을 수행하였다.

Fall accidents require a lot of rehabilitation costs. Since most of the accidents can be prevented compared to other safety accidents, it is more important to prevent them in advance by providing correct information. In particular, in the case of the elderly, body aging and disease due to an increase in age can lead to an increase in safety accidents, and the degree of injury is serious. Therefore, in this paper, we designed predictive data modeling for a fall accident based on the data collected in real time from the integrated sensor module attached to the patient. In addition, data analysis that can monitor falls in real time was studied. To this end, data was collected in real time from temperature sensors, heart rate sensors, and motion sensors, and fall prediction data analysis was performed by applying binomial classification modeling.

최근 많은 게임이 사용자의 게임 플레이와 관련된 데이터를 제공하고 있고, 이에 기계학습 기법을 결합하여 상대의 행동을 예측하는 연구들이 있다. 본 연구는 실시간 전략 게임(클래시로얄)의 경기 데이터와 기계학습 기반의 다 중 레이블 분류를 사용하여 상대 플레이어의 행동을 예측한다. 초기 실험은 이진 형태의 카드 특성과 카드 배치 좌표 그리고 정규화된 시간 정보를 입력받아 카드 타입, 카드 배치 좌표를 랜덤포레스트와 다층 퍼셉트론을 이용하여 예측한 다. 이후, 순차적으로 3 가지 전처리 방식을 사용하여 실험을 진행했다. 먼저 입력 데이터의 특성 정보 일부를 변환 시켜 예측했다. 다음으로 입력 데이터를 연속된 카드 입력 방식까지 고려한 중첩 형태로 변환 시켜 예측했다. 마지막으 로 모든 이전 단계의 데이터들을 정규화된 시간 기준에 따라 초반, 후반으로 분할하여 예측했다. 그 결과 가장 개선을 보인 전처리 방식은 중첩 형태의 데이터를 초반으로 분할하였을 경우로 카드 타입이 약 2.6%, 카드 배치 좌표가 약 1.8% 개선을 보였다.

Recently, many games provide data related to the users' game play, and there have been a few studies that predict opponent move by combining machine learning methods. This study predicts opponent move using match data of a real-time strategy game named ClashRoyale and a multi-label classification based on machine learning. In the initial experiment, binary card properties, binary card coordinates, and normalized time information are input, and card type and card coordinates are predicted using random forest and multi-layer perceptron. Subsequently, experiments were conducted sequentially using the next three data preprocessing methods. First, some property information of the input data were transformed. Next, input data were converted to nested form considering the consecutive card input system. Finally, input data were predicted by dividing into the early and the latter according to the normalized time information. As a result, the best preprocessing step was shown about 2.6% improvement in card type and about 1.8% improvement in card coordinates when nested data divided into the early.