Earticle

연구목적: 본 연구는 렌터카 이용률이 높은 제주특별자치도를 대상으로 렌터카 교통사고 심각도에 영 향을 미치는 주요 요인을 분석하고, 사고감소를 위한 체계적인 방안을 제시하는 것을 목적으로 한다. 이 를 위해 기계학습 기법을 활용하여 사고심각도 분석모델을 개발하였다. 연구방법: 2018년부터 2022년 까지의 제주특별자치도 렌터카 교통사고 데이터를 바탕으로 XGBoost, Decision Tree, Random Forest 를 비교 분석하였다. 최적의 모델로 Random Forest를 선정하고, SHAP 분석을 통해 변수 중요도를 도출 하였다. 연구결과: 랜덤 포레스트 모형이 가장 우수한 성능을 나타냈으며 변수 중요도 분석 결과, 사고 심각도에 큰 영향을 미치는 요인은 사고유형이 ‘차대 보행자’(0.2132169), ‘차대 차’(0.150534) 사고인 경우로 나타났다. 법규 위반 중 ‘안전운전 불이행’(0.145027), ‘중앙선 침범’(0.097202), 도로 형태 ‘교차로’(0.049608)인 경우에도 높은 변수 중요도를 보였다. 또한, 기상 상태_비, 야간, 특정요일(일요일, 목요일), 계절(가을, 여름)에 사고심각도 가 높게 나타났다. 결론: 렌터카 사고감소를 위해 맞춤형 차량 대여 시스템, 교통안전 시설물 확충, 법규 위반 예방조치를 강화해야 하며, 향 후 연구에서는 분석 범위를 확대하고 정책의 효과성을 평가하여 실효성 있는 개선 방안 도출이 필요하다.

Purpose: This study aims to analyze the key factors influencing the severity of rental car crashes in Jeju Special Self-Governing Province, where the rental car usage rate is high, and propose systematic measures to reduce crash severity. To achieve this, a crash severity analysis model was developed using machine learning techniques. Method: Data on rental car crashes in Jeju Special Self-Governing Province from 2018 to 2022 were analyzed using XGBoost, Decision Tree, and Random Forest models. Random Forest was identified as the optimal model, and SHAP analysis was employed to derive the feature importance of variables. Result: The Random Forest model demonstrated the highest performance. The feature importance analysis revealed that the most significant factors influencing crash severity were crash type, with “vehicle-pedestrian collisions”(0.2132169) and “vehicle-to-vehicle collisions”(0.150534) having the highest impact. Among traffic violations, “failure to drive safely” (0.145027) and “center line violations”(0.097202) were identified as critical factors. Additionally, the “intersection” road type (0.049608) was a significant factor. Environmental conditions such as rainy weather, nighttime, specific days(Sunday and Thursday), and seasons (autumn and summer) also increased crash severity. Conclusion: To reduce rental car crash severity, implementing a tailored vehicle rental system, expanding traffic safety infrastructure, and strengthening measures to prevent traffic violations are essential. Future research should expand the analysis scope and evaluate the effectiveness of policies to derive actionable and effective improvement strategies.

불안정한 기후와 함께 나타나는 국지적 집중호우로 인한 도시 침수는 끊임없이 발생하고 있으나, 강우량을 포함한 기상정보 현황 또는 예보정보를 활용하여 공간적인 도시홍수 예측정보를 제공할 수 있는 체계는 아직 마련되지 못한 상황이다. 공간적인 홍수정보는 하천의 제방, 도시 하수관거의 통수능, 저류지, 펌프시설과 같은 구조물적 대책에 어려움이 있을 시 발생할 수 있는 최악의 홍수상황을 미리 파악함으로써 피해를 최소화하는데 직접적인 영향을 미칠 수 있다. 이에 본 연구에서는 기상청에서 제공되는 강수량, 도시 유역에 대한 2차원 침수해석 결과, 그리고 기계학습 모형 중 하나인 랜덤포레스트 회귀모형을 활용하여 실시간으로 도시유역에 대한 침수지도를 예측할 수 있는 방법론을 제시하고자 한다. 연구유역은 내수침수가 빈번하게 발생하는 울산시 우정태화지구로 선정하였다. 지속시간 6시간의 총강우량 50 mm, 80 mm 그리고 110 mm 대한 랜덤포레스트 회귀분석 예측 침수면적과 검보정된 2차원 물리모형의 침수해석 결과 비교시 각각 63%, 80%, 그리고 67%의 적합도를 보여주어, 빠른 시간안에 발생하는 도시 침수에 대한 대응, 대피를 위한 기초자료로 활용될 수 있을 것으로 판단된다.

Urban flooding caused by localized heavy rainfall with unstable climate is constantly occurring, but a system that can predict spatial flood information with weather forecast has not been prepared yet. The worst flood situation in urban area can be occurred with difficulties of structural measures such as river levees, discharge capacity of urban sewage, storage basin of storm water, and pump facilities. However, identifying in advance the spatial flood information can have a decisive effect on minimizing flood damage. Therefore, this study presents a methodology that can predict the urban flood map in real-time by using rainfall data of the Korea Meteorological Administration (KMA), the results of two-dimensional flood analysis and random forest (RF) regression model. The Ujeong district in Ulsan metropolitan city, which the flood is frequently occurred, was selected for the study area. The RF regression model predicted the flood map corresponding to the 50 mm, 80 mm, and 110 mm rainfall events with 6-hours duration. And, the predicted results showed 63%, 80%, and 67% goodness of fit compared to the results of two-dimensional flood analysis model. It is judged that the suggested results of this study can be utilized as basic data for evacuation and response to urban flooding that occurs suddenly.

본 연구는 보(weir)의 기립·전도 운영에 따른 수심 증감률을 정량적으로 예측하기 위해, 가상하도에 대한 사전 모의분석에서는 보면적비(AR)와 하상경사(S)를 이용하여 SPSS 기반 단순선형회귀식을 산정하고, 수리변수의 개별 영향을 파악하여 회귀 구조의 기초를 정립하였다. 실제 하천에서는 3개 하천(만경강·금호강·무심천) 자료를 통합 분석하여 보면적비(AR), 마찰경사(FS), 지점경사(SS) 등을 중심으로 다중선형회귀식(OLS)을 구축하고 랜덤포레스트(RF) 회귀모형과 비교·검증하였다. 랜덤포레스트(RF) 회귀분석을 위해 하천-보 GroupID로 Group K-Fold (K=5) OOF(Out-of-Fold)를 적용하여 학습(≈80%)·검증(≈20%)을 구분하고, 부분의존도(PDP-1D/2D)로 단일·복합 변수의 평균 영향과 상호작용을 해석하였다. 그 결과, 기립은 상호작용 확대형(AR×Fr, AR×FS 등), 전도는 푸르드수(Fr) 포함 단항형이 최적 구조로 선정되었으며, 이를 통해 다중선형회귀식(RF-2D)을 개발하였다. 보면적비(AR)만을 이용하는 RF-AR의 설명력은 제한적이었으나, RF-2D 개발식은 OLS와 유사한 수준의 일반화 성능을 보였다. 본 연구에서 개발된 식은 하천에서 복잡한 수치해석 없이도 제한된 입력자료만으로 보의 기립과 전도에 따른 수심 변화를 신속하고 간편하게 예측할 수 있어, 보 운영 및 하천관리의 사전 의사결정 지원 도구로 활용 가능할 것이다.

This study aims to quantitatively predict the depth-change rate associated with weir operations under raised and lowered conditions. In the preliminary analysis using a virtual channel, a simple linear regression model(SPSS-based) was established using the weir area ratio (AR) and bed slope (S) to examine the individual hydraulic influences and to define the basic structure of the regression model. For the real-river analysis, data from three rivers (Mangyeong, Geumho, and Musim) were integrated to develop a multiple linear regression (OLS) model focusing on AR, friction slope (FS), and site slope (SS), and the results were compared and verified with a Random Forest (RF) regression model. To perform the RF analysis, Group K-Fold (K=5) cross-validation was applied using river-weir GroupID, strictly separating the training (~80%) and validation (~20%) datasets. The partial dependence (PDP-1D/2D) method was used to interpret the mean effects and interactions of single and combined variables. As a result, the raised condition was best represented by an interaction-expanded model (AR×Fr, AR×FS, etc.), while the lowered condition was effectively explained by a single-variable model including Froude Number(Fr). From these results, a multi-variable regression equation(RF-2D) was developed. Although the RF-AR model using only AR showed limited explanatory power, the proposed RF-2D model exhibited a generalization performance comparable to that of the OLS model. The proposed regression equations enable rapid and consistent prediction of water-depth changes due to weir raising or lowering, using only limited input variables and without complex numerical simulations, and can thus serve as a decision-support tool for efficient weir operation and river management.

이 연구의 목적은 서리 발생일과 무상일 기간의 특성을 분석하고 로지스틱 회귀, 랜덤 포레스트, Long-short Term Memory (LSTM) 기법을 활용하여 서리발생 예측모델을 개발하고 평가하는데 있다. 수원, 청주, 광주 지점에서 봄철과 가을철 서리발생 예측모델 개발을 위한 기상변수들을 수집하였으며, 수집기간은 1973년부터 2019년까지이다. 프리시전(precision), 리콜(Recall), f-1 스코어와, AUC 및 Reliability Diagram과 같은 그래피컬 평가기법을 이용해 서리발생 예측모델을 평가하였다. 봄철과 가을철 모두 서리발생일이 줄어드는 경향성(유의수준: 0.01)을 보였다. 0.9 이상의 높은 AUC 값에도 불구하고, 신뢰도는 일정한 값을 보여주지는 않았다. 서리발생일 측뿐만 아니라, 초상일과 종상일을 정확히 예측할 수 있도록 모형 개선이 필요해 보이며, 다른 지역의 더 많은 지점에서 동일한 기법을 적용해 보는 연구가 필요해 보인다.

We investigated changes in frost days and frost-free periods and to comparatively assess frost event prediction models developed using logistic regression (LR), random forest (RF), and long short-term memory (LSTM) networks. The meteorological variables for the model development were collected from the Suwon, Cheongju, and Gwangju stations for the period of 1973-2019 for spring (March - May) and fall (September - November). The developed models were then evaluated by Precision, Recall, and f-1 score and graphical evaluation methods such as AUC and reliability diagram. The results showed that significant decreases (significance level of 0.01) in the frequencies of frost days were at the three stations in both spring and fall. Overall, the evaluation metrics showed that the performance of RF was highest, while that of LSTM was lowest. Despite higher AUC values (above 0.9) were found at the three stations, reliability diagrams showed inconsistent reliability. A further study is suggested on the improvement of the predictability of both frost events and the first and last frost days by the frost event prediction models and reliability of the models. It would be beneficial to replicate this study at more stations in other regions.

본 연구는 비교교육 관점에서 한국과 핀란드 중학교 교사의 효능감에 영향을 미치는 요인 을 비교 분석하는 데 목적이 있다. 이를 위하여 랜덤포레스트 및 더미변수최소자승 회귀분석 (LSDV 모형)을 활용하여 교사효능감에 영향을 미칠 수 있는 다양한 변수들을 종합적으로 탐 색하였다. 아울러 해당 변수들이 한국과 핀란드의 국가적 맥락에 상관없이 공통적 요인인지, 혹은 각 국가의 고유한 특성인지를 살펴 개별 국가의 교원 정책에 관한 정책적 시사점을 도 출하고자 하였다. 한국과 핀란드를 대상으로 교사효능감에 영향을 미치는 요인을 비교 분석 한 결과, 교수활동, 설명의 명확성, 수업 자율성에 대한 만족도 등과 같은 요인이 교사효능감 에 영향을 미치는 공통적인 요인임을 확인하였다. 아울러 교사효능감의 모든 하위영역에서 국가별로 고유하게 나타나는 관련요인이 존재함을 확인하였다. 한국의 경우 학교의 팀 혁신 성이, 핀란드의 경우 다문화 수업의 효능감과 교직만족도가 해당 국가의 고유한 관련요인이 었다. 이런 결과를 토대로 본 연구에서는 교사효능감 관련요인이 국가별 맥락성과 고유성이 존재한다는 점에서 개별 국가의 맥락과 특성을 고려한 정책차용적 관점과 더불어 개별 국가 의 고유한 특성을 적극적으로 반영한 교원정책의 수립을 적극적으로 모색하는 것이 필요하다 는 점을 제언하였다.

This study aims to conduct a comparative analysis of the determinants influencing the self-efficacy of teachers in South Korea and Finland from a comparative education perspective. To achieve this, we employed Random Forest and Least Squares Dummy Variable (LSDV) regression analysis to comprehensively explore various variables that could impact teachers’ self-efficacy. Additionally, we sought to examine whether these variables were common factors across both countries or unique to each nation's context, in order to derive policy implications. The comparative analysis of the factors influencing teachers’ self-efficacy in South Korea and Finland revealed that factors such as instructional activities, clarity of explanations, and satisfaction with instructional autonomy were common factors enhancing teacher self-efficacy. Furthermore, we identified unique factors influencing teacher self-efficacy in each country. In the case of South Korea, team innovativeness within schools was a distinctive factor, while in Finland, the self-efficacy related to multicultural teaching and overall job satisfaction were specific influences. Based on these findings, this study suggests that within the South Korean context, policies aimed at promoting teacher expertise in instructional activities, instructional autonomy, and school-based team innovation are essential, and emphasizes the need to enhance the reciprocal relationship between these influential factors in teacher professional development policies.

Effective nonpoint source (NPS) pollution management requires frequent water quality monitoring, which is, however, often costly to be implemented in practice. Statistical techniques and machine learning methods allow us to identify and focus on fundamental environmental variables that have close relationships with NPS pollutants of interest. This study developed surrogate models to predict the concentrations of suspended sediment (SS) and total phosphorus (T-P) from turbidity and runoff discharge rates using multiple linear regression (MLR) and random forest (RF) methods. The RF models provided acceptable performance in predicting SS and T-P, especially when runoff discharge rates were high. The RF models outperformed the MLR models in all the cases. Such finding highlights the potential of RF techniques and models as a tool to identify fundamental environmental variables that are measured in relatively inexpensive ways or freely available but still able to provide information required to quantify the concentrations of NP S pollutants. The analysis of relative importance rates showed that the temporal variations of SS and T-P concentrations could be more effectively explained by that of turbidity than runoff discharge rate. This study demonstrated that the advanced statistical techniques such as machine learning could help to improve the efficiency of NPS pollutants monitoring.

Wireless communication technology is becoming increasingly prevalent in smart factories, but the rise in the number of wireless devices can lead to interference in the ISM band and obstacles like metal blocks within the factory can weaken communication signals, creating radio shadow areas that impede information exchange. Consequently, accurately determining the radio communication coverage range is crucial. To address this issue, a Radio Environment Map (REM) can be used to provide information about the radio environment in a specific area. In this paper, a technique for estimating an indoor REM using a mobile robot and machine learning methods is introduced. The mobile robot first collects and processes data, including the Received Signal Strength Indicator (RSSI) and location estimation. This data is then used to implement the REM through machine learning regression algorithms such as Extra Tree Regressor, Random Forest Regressor, and Decision Tree Regressor. Furthermore, the numerical and visual performance of REM for each model can be assessed in terms of R2 and Root Mean Square Error (RMSE).

In the quest for advancing diabetes diagnosis, this study introduces a novel two-step machine learning approach that synergizes the probabilistic predictions of Logistic Regression with the classification prowess of Random Forest. Diabetes, a pervasive chronic disease impacting millions globally, necessitates precise and early detection to mitigate long-term complications. Traditional diagnostic methods, while effective, often entail invasive testing and may not fully leverage the patterns hidden in patient data. Addressing this gap, our research harnesses the predictive capability of Logistic Regression to estimate the likelihood of diabetes presence, followed by employing Random Forest to classify individuals into diabetic, pre-diabetic or nondiabetic categories based on the computed probabilities. This methodology not only capitalizes on the strengths of both algorithms—Logistic Regression's proficiency in estimating nuanced probabilities and Random Forest's robustness in classification—but also introduces a refined mechanism to enhance diagnostic accuracy. Through the application of this model to a comprehensive diabetes dataset, we demonstrate a marked improvement in diagnostic precision, as evidenced by superior performance metrics when compared to other machine learning approaches. Our findings underscore the potential of integrating diverse machine learning models to improve clinical decision-making processes, offering a promising avenue for the early and accurate diagnosis of diabetes and potentially other complex diseases.

The model-based evolutionary algorithms are divided into three groups: estimation of distribution algorithms, inverse modeling, and surrogate modeling. Existing inverse modeling is mainly applied to solve multi-objective optimization problems and is not suitable for many-objective optimization problems. Some inversed-model techniques, such as the inversed-model of multi-objective evolutionary algorithm, constructed from the Pareto front (PF) to the Pareto solution on nondominated solutions using a random grouping method and Gaussian process, were introduced. However, some of the most efficient inverse models might be eliminated during this procedure. Also, there are challenges, such as the presence of many local PFs and developing poor solutions when the population has no evident regularity. This paper proposes inverse modeling using random forest regression and uniform reference points that map all nondominated solutions from the objective space to the decision space to solve many-objective optimization problems. The proposed algorithm is evaluated using the benchmark test suite for evolutionary algorithms. The results show an improvement in diversity and convergence performance (quality indicators).

SAR (Synthetic aperture radar) images are less affected by the weather compared to optical images and can be obtained at any time of the day. Therefore, SAR images are being actively utilized for military applications and natural disasters. However, because SAR data are in grayscale, it is difficult to perform visual analysis and to decipher details. In this study, we propose a color mapping method using RF (random forest) regression for enhancing the visual decipherability of SAR images. COSMO-SkyMed 2 and WorldView-3 images were obtained for the same area and RF regression was used to establish color configurations for performing color mapping. The results were compared with image fusion, a traditional color mapping method. The UIQI (universal image quality index), the SSIM (structural similarity) index, and CC (correlation coefficients) were used to evaluate the image quality. The color-mapped image based on the RF regression had a significantly higher quality than the images derived from the other methods. From the experimental result, the use of color mapping based on the RF regression for SAR images was confirmed.

Objectives: The hospital standardised mortality ratio (HSMR) is the ratio of the observed number of hospital deaths tothe expected number of deaths, with the latter estimated using statistical models that adjust for available case-mix factors. This study aimed to develop and validate in-hospital or 30-day post-discharge mortality prediction models for 40 diagnosisgroups within the HSMR framework, using penalized logistic regression (pLR) and random forest (RF), and to compare theperformance of these two approaches. Methods: We analysed 1,144,890 hospital admissions from 14 Malaysian state hospitalsbetween 2012 and 2016. Separate models were developed for each diagnosis group using nine administrative features,including age, comorbidities, and admission category. Model performance was evaluated using mean Brier scores and Cstatisticsacross multiple bootstrapped datasets to obtain less biased performance estimates. Aggregate expected mortalitycounts were also compared with observed counts. Results: The overall observed mortality rate was 10.2%. The pLR modelsconsistently showed better discrimination and calibration than the RF models, with lower Brier scores and higher C-statisticsacross the 40 diagnosis groups. On average, the C-statistic for pLR exceeded that for RF by 0.062. Although the RF modelproduced aggregate mortality predictions that were numerically closer to the observed counts, it showed high variance andpoorer probabilistic calibration than pLR. Conclusions: The pLR model tended to underestimate mortality more than RF butstill demonstrated better calibration and discrimination, making it the preferable model for HSMR analysis in this dataset.

이 연구에서는 랜덤 포레스트 회귀분석을 사용하여 일반대학원에 재학 중인 대학원생의 학습몰입에 주요한 영향을 미치는 설명 변인을 탐색하였다. 분석을 위하여 서울 소재 A 대학의 일반대학원에재학 중인 대학원생 455명의 설문 자료를 사용하였다. 분석에 활용된 설명변수로는 성별, 전공 등 개인 배경 변인 10문항, 지도교수 관련 12문항, 교수학습 및 교육과정의 만족도 15문항, 학사제도, 장학지원 등의 교육 환경지원에 대한 만족도 15문항, 전공문화, 리라밸(Research-Life Balance: 연구-삶균형), 경력개발에 관련한 10문항을 포함하여 총 62문항이 있다. 일반대학원생의 학습몰입을 예측하는주요 설명 변인은 다음과 같이 나타났다. 첫째, 동료지지를 통한 전공 문화와 관련된 4문항은 모두 상위 10위 안의 주요 변수로 나타났다. 둘째, 최신 학술 동향을 반영한 교육과정에 대한 만족도가 주요설명변수였다. 세 번째로, 경력개발과 관련한 인적 네트워크에 대한 만족도가 학습몰입에 주요 변수로 나타났다. 지도교수에 관련한 문항은 상위 30개의 주요 변수에 포함되었으나, 전공문화나 경력개발보다는 중요도가 떨어졌다. 분석에 투입된 설명변수의 상대적 중요도를 바탕으로 대학원생의 학습몰입을 예측하는 주요 변수를 도출하였으며, 연구 결과를 통해 대학원 교육의 질 증대를 위한 실무적방안을 제안하였다.

The purpose of the study was to examine influential factors on graduate students’ learning engagement using random forest regression (RFR). For analysis, this study used 455 responses to a graduate school satisfaction survey at one university. The explanatory variables for RFR consisted of 10 individual variables, including gender and major; 12 supervisor-related items; 15 educational curriculum and teaching items; 15 educational environment support items; 4 co-worker support items; 3 social networks for career development items; and 3 research-life balance items. In summary, 62 items were used as the exploratory variables. The results of the study were as follows. First, the results showed that co-worker support for all four items was the most important explanatory variable. Second, the educational curriculum item involving a recent research trend was an important influential factor. Third, social networks for career development were considered a significant explanatory variable. Although the graduate education system is based on supervisor apprenticeship, supervisor factors were considered less important than co-worker support. Finally, this study suggests policy implications for improving the quality of graduate school education.

본 연구는 청소년의 자아존중감 관련 요인을 탐색하고 변수 간 영향의 방향성과 비선형적 관계를 분석하기 위해랜덤포레스트 회귀모형을 활용하였다. 연구 자료는 한국청소년정책연구원의 「10대 청소년의 정신건강 실태조사」를활용하였으며, 전국 17개 시도의 청소년 6,689명을 대상으로 분석을 수행하였다. 자아존중감을 결과변수로 설정하고, 심리·가족·학교·건강·인구사회학적 요인을 포함한 다양한 변수를 동시에 투입하였다. 분석은 Permutation Importance, SHAP, Partial Dependence Plot을 활용하여 수행하였다. 연구 결과, 외모만족도와 스트레스가 자아존중감에 가장 밀접하게 관련된 변수로 나타났으며, 우울, 부모양육(따스함), 가족관계, 학교적응 관련 변수들도 상대적으로 높은 중요도를 보였다. SHAP 분석 결과 외모만족도는 자아존중감과 정적인 관련성을 보이는 반면, 스트레스와 우울은 자아존중감을 부적인 관련성으로 작용하였다. 또한 PDP 분석에서는 주요 변수들이 자아존중감과 비선형적 관계를 보였으며, 특히스트레스와 우울은 일정 수준 이후 자아존중감이 급격히 감소하는 경향이 확인되었다. 본 연구는 청소년 자아존중감향상을 위해 긍정적 신체상 형성, 스트레스 관리, 관계적·환경적 지지 강화가 중요함을 시사한다.

This study employed a Random Forest regression model to explore factors associated Bith adolescent self-esteem and to examine their nonlinear relationships. Data Bere draBn from the Mental Health Survey of Adolescents conducted by the National Youth Policy Institute, including 6,689 adolescents across South Korea. Self-esteem Bas the outcome variable, and psychological, familial, school-related, health-related, and sociodemographic factors Bere included as predictors. Permutation Importance, SHAP, and Partial Dependence Plot (PDP) analyses Bere conducted. The results indicated that appearance satisfaction and stress Bere the factors most strongly associated Bith self-esteem, folloBed by depression, parental Barmth, family relationships, and school adjustment. The SHAP analysis shoBed positive associations betBeen appearance satisfaction and self-esteem and negative associations for stress and depression. PDP analysis further revealed nonlinear relationships, Bith self-esteem declining sharply beyond certain levels of stress and depression. These findings highlight the importance of positive body image, stress management, and supportive environments in promoting adolescent self-esteem.

본 연구는 머신러닝 기법의 하나인 랜덤 포레스트 회귀 방법을 적용, 청소년의 정신건강 문제를 예측하는 요인을 탐색하여 이를 바탕으로 청소년의 정신건강 증진을 위한 기초자료를 제공하고자 하였다. 이를 위해 경기교육종단연구(GEPS) 9차년도 고등학생 5,183명의 자료를 사용하였으며, 주요 예측변인 30개를 중요도 지수 기준으로 도출하였다. 주요 연구 결과는 다음과 같다. 청소년의 정신건강 문제를 예측하는 주요 요인은 자아존중감, 부모와의 관계(소외감), 학습 무동기 및 부과된 조절동기, 시험 및 학업 스트레스, 성별, 신체적 자아개념, 고정신념, 교과 흥미 및 효능감, 스마트폰 사용 시간 등인 것으로 밝혀졌다. 부모 및 학교 수준의 변수 중 유의한 영향 요인은 없는 것으로 나타났다. 이와 같은 연구 결과를 바탕으로 청소년의 정신건강 문제를 개선할 수 있는 시사점을 논의하였다.

This study aimed to explore the predictors of adolescent mental health issues by applying the Random Forest regression technique, a machine learning method, to provide foundational data for improving adolescent mental health. Data from the 9th year of the Gyeonggi Education Longitudinal Study (GEPS) involving 5,183 high school students were utilized, and 30 key predictive variables were identified based on importance indices. The main findings are as follows: Significant predictors of adolescent mental health problems included self-esteem, feelings of alienation in parental relationships, academic disengagement and imposed regulatory motivation, exam and academic stress, gender, physical self-concept, fixed mindset, interest and efficacy in academic subjects, and smartphone usage time. No significant effects were found for variables at the parental or school levels. Based on these results, implications for improving adolescent mental health issues are discussed.

본 연구에서는 랜덤 포레스트 회귀(random forest regression)를 활용하여 중·고등학교 1학년의 교우관계를 예측하는 영향요인을 개별 문항 단위에서 분석하였다. 이를 위해 경기교육종단연구(GEPS) 4, 7차년도 학생, 학부모 자료를 사용하였다. 연구 결과는 다음과 같다. 첫째, 중학교 1학년 교우관계 영향요인 상위 30개는 교사-학생관계, 자아개념-사회, 시민의식, 자아존중감, 희망대학전공, 진로자기이해, 직업태도, 진학희망고교, 교육포부, 국어와 영어성취도, 스마트폰사용빈도, 가구소득, 성별, 차별, 학업스트레스, 방학생활, 전반적만족 영역에 포함되는 문항들이었다. 둘째, 고등학교 1학년 교우관계 영향요인 상위 30개는 교사-학생관계, 자아개념-사회, 자아존중감, 학업스트레스, 학습자기이해, 학습전략-조직화, 희망대학전공, 고등학교 진학희망이유, 직업선택기준, 직업태도, 애착-신뢰, 방학생활, 동아리 영역에 포함되는 문항들이었다. 영향요인이 속한 영역이 동일하더라도 해당 영역에서 영향력이 있는 것으로 선별된 개별 문항은 학교급에 따라 달랐다. 공통 영향요인 영역은 교사-학생관계, 자아개념-사회, 희망대학전공, 직업태도, 학업스트레스, 방학생활이었다.

The purpose of this study is to examine the influence factors of peer relations among the first graders of middle and high school with Random Forest regression. This study used the GEPS data. The subjects of the study were 5,591 middle school students and 5,339 high school students. The results were as follows: First, the main factors among peer relations in the first graders of middle school were included in the ‘teacher-student relations’, ‘self concept-society’, ‘civic consciousness’, ‘self-esteem’, ‘desired major’, ‘career self-understanding’, ‘job attitude’, ‘desired high school’, ‘educational aspiration’, ‘achievement of Korean and English’, ‘frequency of smartphone use’, ‘family income’, ‘gender’, ‘discrimination’, ‘academic stress’, ‘life during vacation’, and ‘overall satisfaction’. Second, the main factors among peer relations in the first graders of high school were included in the ‘Teacher-student relations’, ‘self concept-society’, ‘self-esteem’, ‘academic stress’, ‘learning self-understanding’, ‘learning strategy-organization’, ‘desired major’, ‘reasons for entering high school’, ‘job selection criteria’, ‘job attitude’, ‘attachment-trust’, ‘life during vacation’, and ‘club activities’. Even if the constructs were the same, the influencing factors of each item were composed differently according to the school level.

본 연구는 국내 소년범 재범 예측 진단 체계의 한계점을 비판적으로 고찰하고 머신러닝 기반의 랜덤 포레스트(Random Forest) 알고리즘을 적용하여 소년범 재범위험 예측 모형의 개발 및 검증을 시도하였다. 기존 소년범 재범 예측 도구는 주로 요인분석 또는 로지스틱 회귀분석을 기반으로 개발되어 왔으나 변수 간 비선형적 상호작용과 데이터 불균형, 과적합 등 다양한 한계를 지닌다. 본 연구에서는 법무부 소년원 출원생 1,736명의 TEAMS 자료를 분석 대상으로 하여 로지스틱 회귀와 랜덤 포레스트 모형의 예측 성능을 교차검증 및 오버샘플링(SMOTE), 임계값 최적화 등의 절차를 통해 비교ㆍ평가하였다. 그 결과, 랜덤 포레스트 모형은 로지스틱 회귀보다 일관되게 높은 AUC를 보이며 우수한 예측력을 나타냈으나, 민감도-특이도 간 불균형 문제가 존재하였고, 최적 임계값 적용을 통해 일정 부분 개선 가능함을 확인하였다. 또한 변수 중요도 및 부분의존도 분석을 통해 연령, 누적 비행력 등 주요 예측 변수를 도출하였다. 본 연구는 XAI(설명가능 인공지능) 및 SHAP 분석의 필요성을 논의하며 향후 맞춤형 처우와 과학적 사례 관리, 연구 윤리 확립 및 실증적 연구 문화 정착의 정책적ㆍ학문적 함의를 제시하였다.

This study critically examines the limitations of the current risk assessment systems for predicting juvenile recidivism in South Korea and seeks to develop and validate a machine learning-based prediction model using the Random Forest algorithm. Existing tools for juvenile recidivism prediction have been primarily developed using factor analysis or logistic regression, which face various limitations such as inability to capture nonlinear variable interactions, data imbalance, and overfitting. Using TEAMS data from 1,736 juvenile releasees from juvenile correctional institutions, this study compares and evaluates the predictive performance of logistic regression and Random Forest models through cross-validation, oversampling (SMOTE), and optimal cutoff value adjustment. The results show that the Random Forest model consistently outperforms logistic regression in terms of AUC, demonstrating superior predictive power, although issues with sensitivity-specificity imbalance remain. These were partly addressed by applying optimal cut-off points. Variable importance and partial dependence analyses identified age and cumulative delinquency as key predictors. The study discusses the necessity of explainable AI (XAI) and SHAP analyses, and highlights the policy and academic implications for tailored interventions, scientific case management, the establishment of research ethics, and the advancement of empirical research culture.

Flow-Accelerated Corrosion (FAC) is a phenomenon in which a protective coating on a metal surface is dissolved by a flow of fluid in a metal pipe, leading to continuous wall-thinning. Recently, many countries have developed computer codes to manage FAC in power plants, and the FAC prediction model in these computer codes plays an important role in predictive performance. Herein, the FAC prediction model was developed by applying a machine learning method and the conventional nonlinear regression method. The random forest, a widely used machine learning technique in predictive modeling led to easy calculation of FAC tendency for five input variables: flow rate, temperature, pH, Cr content, and dissolved oxygen concentration. However, the model showed significant errors in some input conditions, and it was difficult to obtain proper regression results without using additional data points. In contrast, nonlinear regression analysis predicted robust estimation even with relatively insufficient data by assuming an empirical equation and the model showed better predictive power when the interaction between DO and pH was considered. The comparative analysis of this study is believed to provide important insights for developing a more sophisticated FAC prediction model.

본 연구는 정기적 기부 예측을 위해 랜덤 포레스트 머신러닝 알고리즘과 단계별 로지스틱 회귀분석 기법을 활용하였다. 본 연구에서 활용한 데이터는 아름다운재단 기부문화연구소 기빙코리아 설문조사 2020년과 2022년 그리고 합동 횡단면 자료를 활용하였다. 본 연구는 선행연구를 토대로 기부자 특성과 내재적 요인 그리고 외재적 요인으로 구분하여 다양한 투입 변수를 활용하였다. 내재적 요인은 기부 동기를 고려하였으며, 외재적 요인은 비영리단체 운영에 관한 평가 요인으로 반응성, 투명성, 활동성과 그리고 신뢰를 포함하였다. 본 연구 결과에서는 랜덤 포레스트 분석과 로지스틱 회귀분석 모두 정기적 기부 예측에 기부자 특성뿐만 아니라 다양한 내재적 요인과 외재적 요인이 중요변수로 나타났다. 추가로 두 분석에서는 종교 유무와 종교 신뢰가 공통된 중요변수로 나타났으나, 가구원 수와 월 평균 소득은 랜덤 포레스트 분석에서만 중요변수로 나타나는 차이점을 보였다. 이를 바탕으로 본 연구는 다음과 같은 함의를 가진다. 첫째, 비영리단체는 정기적 기부 예측에 필요한 다양한 변수를 탐색할 필요가 있다. 둘째, 비영리단체는 투명성을 높이기 위해 외부감사 활용이 필요하며, 정부 차원에서는 비영리단체에 관한 통합적이며 일관적인 관리체계 구축이 필요하다. 셋째, 협력적 거버넌스 차원에서는 다양한 주체가 자원봉사와 같은 나눔에 대한 체험과 교육 프로그램 개발로 기부문화를 조성하는 노력이 필요하다.

This purpose of this study compare logistic regression analysis and random forests machine learning algorithms for predicting sustainable giving behavior. Previous studies have shown varying results depending on the variables involved. We aim to determine which variables are good predictors of sustainable giving behavior. Therefore, we consider various donor aspects such as intrinsic motivation and nonprofit organization(NPO) evaluation aspects, such as extrinsic motivation. We identified 40 variables for prediction, and we compare logistic regression analysis and random forest analysis across three models: the 2020 model, the 2022 model, and the pooled cross-sectional model. According to the results, personal characteristics, intrinsic motivation, and extrinsic motivation emerge as important variables across both analyses. Common variables identified in both analyses include being religious and trust in religion, while household size and income are identified as differing variables. In particular, household size does not exhibit a consistent linear relationship with predictions but shows dynamics. The results of this study help nonprofit organizations understand the characteristics of donors and their needs, so as to better manage continuous donations in the future.

목적 본 연구는 대학 신입생의 학업 부진에 영향을 미치는 요인을 통합적으로 분석하여 신입생의 학업 적응을 위한 대학 교육 정책연구에 시사점을 제공하고자 한다. 방법 이를 위하여 한국고용패널에서 조사한 4차년도 설문에 응답한 총 8,516명의 대학 신입생 자료 중 학업부진을 판단할 수 있는성적이 기록된 4,454명의 응답을 분석에 사용하였다. 학업부진요인에 대한 다양한 관점에서의 분석을 위해 머신러닝 기법인 Lasso 회귀분석과 랜덤포레스트 분석을 활용하였다. 결과 Lasso 회귀분석 결과, 대학 신입생의 학업 부진에 영향을 미치는 변인은 성별, 계획수행, 기업 채용정보 제공 등이 영향을 미치는 것으로 나타났으며, 랜덤 포레스트 분석 결과 개인의 생활 문화, 가족 배경, 학교 환경 등이 주요 요인으로 나타났다. 새롭게 도출된 변인에 대해 살펴본 결과, 디지털 기술 활용 능력, 학습 습관 및 동기, 정서적 안정감이 확인되었다. 이러한 변인들을 종합적으로고려함으로써 학업 부진에 대한 보다 정확한 이해를 도모할 수 있다. 결론 대학 신입생의 학업 부진에 영향을 미치는 요인은 보다 복잡하고 다양한 것으로 나타났다. 학업 부진을 예방하기 위해 개별화된맞춤형 지원과 디지털 네이티브 세대에 맞는 대학 교육의 혁신이 필요할 것이다.

Objectives This study aims to provide implications for university education policy research for the academic adjustment of freshmen by comprehensively analyzing the factors that affect academic underachievement of freshmen. Methods For this purpose, we used data from a total of 8,516 college freshmen who responded to the fourth-year survey conducted by the Korean Employment Panel, and applied machine learning techniques, namely Lasso regression analysis and Random Forest analysis. Results As a result of Lasso regression analysis, it was found that gender, implementation of the plan, and provision of company recruitment information were the variables that affected academic underachievement of college freshmen. As a result of random forest analysis, personal life culture, family background, and school environment were identified as major factors. As a result of examining the newly derived variables, the ability to use digital technology, learning habits and motivation, and emotional stability were confirmed. By comprehensively considering these variables, a more accurate understanding of academic underachievement can be sought. Conclusions The factors influencing academic underachievement of college freshmen are more complex and diverse. Individualized support and innovation in university education suitable for the digital native generation will be needed to prevent academic underachievement.

The COVID-19 virus appeared in 2019 and is extremely contagious. Because it is very infectious and has a huge impact on people's mobility. In this paper, multiple linear regression and random forest models are used to predict the number of COVID-19 cases using COVID-19 infection status data (open source data provided by the Ministry of health and welfare) and Google Mobility Data, which can check the liquidity of various categories. The data has been divided into two sets. The first dataset is COVID-19 infection status data and all six variables of Google Mobility Data. The second dataset is COVID-19 infection status data and only two variables of Google Mobility Data: (1) Retail stores and leisure facilities (2) Grocery stores and pharmacies. The models' performance has been compared using the mean absolute error indicator. We also a correlation analysis of the random forest model and the multiple linear regression model.