Earticle

사물인터넷 환경에서 인간의 개입 없는 지능화된 서비스를 위해서는 IoT 디바이스에서 생성되는 빅데이터로 부터 정상 패턴을 학습하고 이를 기반으로 불량, 오작동과 같은 이상 징후에 대해 예측하는 과정이 요구된다. 본 연구 의 목적은 제품 공정의 다양한 기기에서 발생되는 빅데이터를 분석함으로써 제품 불량을 예측할 수 있는 기계 학습 모델을 구현하는 것이다. 기계 학습 모델은 어느 정도 볼륨을 가진 기존 데이터를 기반으로 분석을 해야 하므로 빅데 이터 분석도구 R을 사용하였으며, 제품 공정에서 수집된 데이터에는 제품에 대한 불량 여부가 포함되어 있으므로 지도 학습 모델을 활용하였다. 연구의 결과, 제품 불량에 영향을 주는 변수 및 변수 조건을 분류하였고, 의사결정 트리를 기반으로 제품의 불량 여부에 대한 예측 모델을 제시하였다. 또한, ROC Curve를 이용한 모델의 적합성 및 성능평가 분석에서 모델의 예측력은 상당히 높게 나타났다.

In order to provide intelligent services without human intervention in the Internet of Things environment, it is necessary to analyze the big data generated by the IoT device and learn the normal pattern, and to predict the abnormal symptoms such as faulty or malfunction based on the learned normal pattern. The purpose of this study is to implement a machine learning model that can predict product failure by analyzing big data generated in various devices of product process. The machine learning model uses the big data analysis tool R because it needs to analyze based on existing data with a large volume. The data collected in the product process include the information about product faulty, so supervised learning model is used. As a result of the study, I classify the variables and variable conditions affecting the product failure, and proposed a prediction model for the product failure based on the decision tree. In addition, the predictive power of the model was significantly higher in the conformity and performance evaluation analysis of the model using the ROC curve.

본 연구에서는 그래프 신경망 모형을 포함한 6개의 기계학습모형을 활용하여 한국 기업의 신용변동을 예측했다. 이를 예측하기 위해 수익성, 성장성, 안정성, 활동성 지표들을 활용하였고, 주성분 분석을 진행하여 지표별로 두 개씩 활용했다. 또한, 신용등급변동이 없는 경우가 대다수인 불균형 데이터를 처리하기 위해 SMOTE 방법 론을 활용하여 데이터를 오버샘플링했다. 본 연구에서는 산업구조를 반영하기 위해 이를 나타내는 상장시장, 데이터 시점, 현재 신용등급과 업종의 네 분류로 나누어 모형을 만들고 이를 최종적으로 결합하여 예측 결과를 도출했다. 결과를 살펴보면, 모든 모형이 데이터 비중에 맞게 선택하는 벤치마크 모형보다 우수하고, 그래프 신경 망 모형이 특히 더 우수한 성능을 나타내는 것을 확인할 수 있다. 또한, 변수 중요도를 분석한 결과, 성장성 지표와 활동성 지표가 중요한 것을 확인할 수 있었고, 그래프 신경망에서는 업종 그래프와 상장시장 그래프가 성능이 좋은 것을 확인할 수 있었다.

Credit rating is an essential component of the financial sector, as it evaluates a firm’s capacity to meet debt obligations. In the Republic of Korea, rating agencies assign levels ranging from AAA to D, with additional modifiers, and these ratings significantly affect financing conditions. Traditional methods typically rely on logistic regression and selected financial variables, yet these approaches often face difficulties in capturing the intricate or nonlinear patterns present in corporate financial data. In response to these challenges, researchers have increasingly turned to advanced machine learning algorithms that can account for more complex relationships. Nevertheless, their deployment is limited by relatively small datasets—particularly among smaller firms—and by concerns regarding model interpretability. The present study proposes a machine learning framework, including a Graph Neural Network (GNN), to predict rating changes in Korean firms. The data set spans 2010 to mid-2024 and includes 182 firms, yielding 1,417 year-level samples. Each annual observation is labeled according to whether its credit rating was upgraded (+1), downgraded (–1), or left unchanged (0). Because most observations lie in the unchanged category, the data are highly imbalanced. To address this imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) is employed, generating additional samples in the minority classes. In addition, Principal Component Analysis (PCA) is utilized to reduce the dimensionality of sixteen indicators representing profitability, growth, stability, and activity. Six models are assessed: logistic regression, LASSO, random forest, support vector machine, Light Gradient Boosting Machine (LGBM), and a GNN. The GNN-based approach is noteworthy for modeling financial indicators or corporate attributes as nodes within a graph, with edges delineating the relationships among these nodes. Such a representation allows for the capture of latent dependencies that are difficult to detect in methods that treat predictors independently. Furthermore, each corporate sample belongs to discrete subgroups determined by listing market (KOSPI or KOSDAQ), data period, rating band, and industry classification. The proposed hierarchical GNN merges these subgroup-specific networks to produce a consolidated prediction, thereby incorporating both firm-level attributes and group-level characteristics. When compared to a benchmark that classifies samples randomly in proportion to the observed class distribution, all six machine learning algorithms demonstrate superior performance. The GNN shows the highest precision and F1-scores, suggesting that it is particularly effective at identifying upgrades and downgrades, which are far less common than no rating changes. Nonetheless, like other models, it finds rare rating shifts more challenging to predict, highlighting the impact of data imbalance and the difficulty of forecasting uncommon events. An inspection of feature importance across models underscores the significance of growth and activity metrics, implying that sales expansion, equity growth, and the efficient use of assets offer robust signals of rating volatility. Moreover, the GNN indicates that distinguishing firms by industry group is especially influential, possibly because each sector’s distinctive regulatory, economic, and financial traits shape its credit risk profile. Compared to certain deep neural networks that demand extensive datasets, the GNN-based method presented here is relatively more practical in settings with limited data, including smaller firms with incomplete rating histories. Additionally, this approach provides improved transparency, as the graph architecture clarifies how different financial indicators or subgroups collectively affect rating transitions. Future work may benefit from enlarging the dataset, experimenting with alternative oversampling strategies such as ADASYN, and examining cost-sensitive learning to mitigate the imbalance problem further. Investigations might also consider alternative graph structures that connect entire firms as nodes and delineate inter-firm relationships or incorporate advanced architectures such as Transformers or LSTM networks. In summary, the findings suggest that a GNN-based framework can improve credit rating predictions by capturing complex interactions that traditional or other advanced machine learning methods may overlook. While data imbalance is still a problem, its consequences are somewhat mitigated by SMOTE. The significance of growth, activity, and sector-specific characteristics suggests that more accurate and comprehensible rating projections can be produced by integrating richer and more interconnected data. In the end, more investigation and more extensive data gathering should improve the precision and dependability of credit rating systems, leading to a better comprehension of the dynamics of corporate finance.

Purpose: This study aimed to develop a machine learning (ML)-based predictive model for hospitallength of stay incorporating clinical, nursing, and healthcare system factors to optimize hospitalresource allocation, improve patient-centered care, and enhance nursing workflow efficiency. Methods: This retrospective study analyzed a large dataset of inpatient electronic medical records froma private tertiary hospital. The dataset was used to develop predictive models for long-term versusshort-term hospitalization. The modeling process involved several ML algorithms, and their performancewas evaluated using standard statistical metrics. The most significant predictive variables wereidentified through an analysis of their feature importance. Results: Among the tested models, the Random Forest algorithm exhibited the highest predictive accuracy,demonstrating strong performance in predicting hospital length of stay. Key influencing factorsincluded the number of consultations, postoperative recovery time, duration of stay in the intensive careunit, the use of third-generation antibiotics, and the need for infection isolation. Patients requiringventilator care, intensive care unit admission, and specific powerful antibiotics were more likely toexperience prolonged hospitalization. Additionally, nursing-related factors such as fall risk and pressureulcer risk were significantly correlated with an extended hospital stay. Conclusion: This study demonstrates that ML models can effectively predict hospital length of stay,aiding in hospital resource management, nursing workforce allocation, and patient safety interventions. The integration of predictive analytics into healthcare systems can support early risk assessment,personalized discharge planning, and overall hospital efficiency.

Purpose: The purposes of this study were to develop a prediction model for pressure injury using a machine learning algorithm and to integrate it into clinical practice. Methods: This was a retrospective study of tertiary hospitals in Seoul, Korea. It analyzed patients in 12 departments where many pressure injuries occurred, including 8 general wards and 4 intensive care units from January 2018 to May 2022. In total, 182 variables were included in the model development. A pressure injury prediction model was developed using the gradient boosting algorithm, logistic regression, and decision tree methods, and it was compared to the Braden scale. Results: Among the 1,389,660 general ward cases, there were 451 cases of pressure injuries, and among 139,897 intensive care unit cases, there were 297 cases of pressure injuries. Among the tested prediction models, the gradient boosting algorithm showed the highest predictive performance. The area under the receiver operating characteristic curve of the gradient boosting algorithm's pressure injury prediction model in the general ward and intensive care unit was 0.86 (95% confidence interval, 0.83~0.89) and 0.83 (95% confidence interval, 0.79~0.87), respectively. This model was integrated into the electronic health record system to show each patient's probability for pressure injury occurrence, and the risk factors calculated every hour. Conclusion: The prediction model developed using the gradient boosting algorithm exhibited higher performance than the Braden scale. A clinical decision support system that automatically assesses pressure injury risk allows nurses to focus on patients at high risk for pressure injuries without increasing their workload.

With the success of the digital economy and the rapid development of its technology, network security has received increasing attention. Intrusion detection technology has always been a focus and hotspot of research. A hybrid model that combines particle swarm optimization (PSO) and kernel extreme learning machine (KELM) is presented in this work. Continuous-valued PSO and binary PSO (BPSO) are adopted together to determine the parameter combination and the feature subset. A fitness function based on the detection rate and the number of selected features is proposed. The results show that the method can simultaneously determine the parameter values and select features. Furthermore, competitive or better accuracy can be obtained using approximately one quarter of the raw input features. Experiments proved that our method is slightly better than the genetic algorithm-based KELM model.

Purpose: This study aimed to develop and validate a machine learning-based prediction model for early hospital readmission (EHR) post-kidney transplantation. Methods: The study was conducted at the organ transplantation center of a university hospital, utilizing data from 470 kidney transplant recipients. We built and trained four machine learning models and tested them to identify the strongest EHR predictors. Predictive performance was evaluated using confusion matrices and the area under the receiver operating characteristic curve (ROC AUC). Results: Among the 470 kidney transplant recipients with a mean age of 46.1±15.30 years, 322 (68.5%) were males, and 74 (15.7%) were readmitted within 30 days after kidney transplantation. In total, 241 (51.2%) recipients were found to have experienced EHR after applying the random over-sampling examples method. The random forest model achieved the best performance, with an ROC AUC of .87 (validation set) and .82 (test set). The 15 most important features were steroid pulse therapy (recipient), cerebrovascular accident (recipient), heart failure (recipient), male sex (donor), cardiovascular disease (recipient), weekend discharge (recipient), peritoneal dialysis (recipient) cerebrovascular accident as the cause of brain death (donor), current smoker (recipient), cardiac arrest (donor), previous kidney transplantation (recipient), age (donor), hypertension (donor), male sex (recipient), and dialysis duration (recipient). Conclusion: Our framework demonstrated strong predictive interpretability. It can support appropriate and effective clinical decision-making by assisting transplant professionals in stratifying recipients based on their risk of EHR. prioritizing post-discharge care and follow-up for high-risk individuals, and allocating targeted interventions such as closer monitoring or education.

Based on the concept of Industry 4.0, various sensors are attached to facilities and equipment to collect data in real time and diagnose faults using analyzing techniques. Diagnostic technology continuously monitors faults or performance degradation of facilities and equipment in operation and diagnoses abnormal symptoms to ensure safety and availability through maintenance before failure occurs. In this paper, we propose a model to analyze the data and diagnose the state or failure using machine learning. The diagnosis model is based on a support vector machine (SVM)-based diagnosis model and a self-learning one-class SVM-based diagnostic model. In the future, it is expected that this model can be applied to facilities used in the entire industry by applying the actual data to the diagnostic model proposed in this paper, conducting the experiment, and verifying it through the model performance evaluation index.

Purpose: This study aimed to develop a predictive model for the early identification of patients at risk of sepsis, using routinely available clinical information and laboratory test results collected during the initial phase of patient care. Methods: This retrospective analysis included electronic medical records of 22,400 adult patients who presented with suspected infection to a tertiary care university hospital in Korea between January 2013 and May 2024. Patients were classified according to Systemic Inflammatory Response Syndrome (score ≥ 2) or Quick Sequential Organ Failure Assessment (score ≥ 2), in combination with sepsis-related International Classification of Diseases, 10th revision codes. Four different machine learning models were trained and validated using five-fold cross-validation. In addition, Shapley additive explanations analysis was performed to interpret the contribution and clinical relevance of key predictive variables. Results: Among the evaluated models, CatBoost demonstrated the strongest predictive performance. Notably, platelet distribution width, alveolar?arterial oxygen difference, procalcitonin, and the arterial/alveolar oxygen ratio consistently emerged as major predictors. Importantly, several variables that did not reach statistical significance in univariate analysis nevertheless contributed substantially to overall model performance, highlighting the importance of complex, multidimensional interactions among clinical factors. Conclusion: These findings indicate that a model based on simple, routinely collected clinical data can achieve high predictive accuracy and strong generalizability. Such a tool may support early clinical decision-making by multidisciplinary teams, including nurses, across diverse real-world care settings. Further prospective studies are warranted to validate its clinical utility and to assess its potential effects on patient outcomes.

Objective To present new classification methods of knee osteoarthritis (KOA) using machine learning and compare its performance with conventional statistical methods as classification techniques using machine learning have recently been developed. Methods A total of 84 KOA patients and 97 normal participants were recruited. KOA patients were clustered into three groups according to the Kellgren-Lawrence (K-L) grading system. All subjects completed gait trials under the same experimental conditions. Machine learning-based classification using the support vector machine (SVM) classifier was performed to classify KOA patients and the severity of KOA. Logistic regression analysis was also performed to compare the results in classifying KOA patients with machine learning method. Results In the classification between KOA patients and normal subjects, the accuracy of classification was higher in machine learning method than in logistic regression analysis. In the classification of KOA severity, accuracy was enhanced through the feature selection process in the machine learning method. The most significant gait feature for classification was flexion and extension of the knee in the swing phase in the machine learning method. Conclusion The machine learning method is thought to be a new approach to complement conventional logistic regression analysis in the classification of KOA patients. It can be clinically used for diagnosis and gait correction of KOA patients.

Purpose: This study aimed to develop a machine learning model to predict exclusive breastfeeding during the first 3 months after birth and to explore factors affecting breastfeeding outcomes. Methods: Data from 2,579 participants in the Korean Early Childhood Education & Care Panel between March 1 and June 3, 2025 were analyzed using Python version 3.12.8 and Colab. The dataset was split into training and testing sets at an 80:20 ratio, and five classifiers (random forest, logistic regression, decision tree, AdaBoost, and XGBoost) were trained and evaluated using multiple performance metrics and feature importance analysis. Results: The confusion matrix of the random forest classifier model demonstrated strong performance, with a precision of 86.6%, accuracy of 84.8%, recall of 96.8%, F1-score of 91.9%, and an area under the curve of 86.0%. Twenty-one features were analyzed, from which feeding plan, breastfeeding at 1 month, marriage period, maternal prenatal weight, self-respect, alcohol consumption, grit, value placed on children, maternal age, and depression emerged as important predictors of exclusive breastfeeding in the first 3 months. Discussion: A robust model was developed to predict exclusive breastfeeding that identified feeding planning and breastfeeding at 1 month as the most influential predictors. The model could be implemented in clinical and community settings to guide tailored breastfeeding support strategies, coupled with the integration of maternal self-respect, grit, and the value placed on children in counseling programs to promote exclusive breastfeeding.

당뇨병은 혈당 수치가 오랜 시간 지속되는 대사질환이다. 당뇨병은 자체 질환에 대한 위험성과 더불어 심혈관과 뇌혈 관, 망막병증, 족부병변 등 질환 다양한 합병증을 유발하는 질병이다. 이렇게 발생하는 합병증으로 인해 사망률을 높이는 원인이 되고 있다. 이에 따라 당뇨병 조기발견과 합병증에 대한 치료가 수반되어야하며, 이를 위해 빠른 진단이 필요하 다. 이에 본 연구에서는 당뇨병 환자 데이터를 이용하여 인공지능 기법 중 로지스틱 회귀분석 머신러닝 알고리즘을 이용한 당뇨병 분류에 대한 머신러닝 모델을 구축하고자 한다. 머신러닝 모델은 scikit-learn을 이용하여 학습용 데이터 80%, 검증용 데이터 20%로 구분하고 교차검증을 5로 나누어 모델을 구성하였다. 로지스틱 회귀분석을 이용한 머신러닝 결과 정확도는 평균 92.24%, AUC 0.95, Recall 0.9294, Precision 0.927, F1 score 0.9248, Kappa 0.6983 이었으 며, 허리-엉덩이 둘레비율이 당뇨병 분류에 가장 큰 영향을 주는 변수였다. 간단한 모델을 통해 높은 성능의 모델을 구축할 수 있었으며, 개인 생활습관 등에 관한 변수를 추가한다면, 더욱 우수한 성능의 머신러닝 모델을 구축할 수 있을 것으로 사료된다.

Diabetes is a metabolic disease in which blood sugar levels last for a long time. Diabetes is a disease that causes various complications of diseases such as cardiovascular and cerebrovascular, retinopathy, and diabetes foot ulcer in addition to the risk of its own disease. Complications that occur in this way are responsible for increasing the mortality rate. Accordingly, early detection of diabetes and treatment for complications should be accompanied, and for this, rapid diagnosis is required. In this study intends to build a machine learning model for diabetes classification using logistic regression machine learning algorithm among artificial intelligence techniques using diabetes patient data. Machine learning model was divided into 80% of training data and 20% of verification data using scikit-learn, and the cross-validation was divided into 5 to construct a model. Accuracy of machine learning results using logistic regression analysis was 92.24% on average, AUC 0.95, Recall 0.9294, Precision 0.927, F1 score 0.9248, and Kappa 0.6983, and the waist-hip circumference ratio was the most influential variable for diabetes classification. A high-performance model could be built through a simple model, and if variables related to personal lifestyle are added, it is believed that a better performance machine learning model can be built.

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%.

Heart disease is a major cause of mortality in the world that is in dire need of accurate, interpretable predictive measures that could be utilized to manage it proactively. The writer of this paper proposes an Explainable AI (XAI) Ensemble Machine Learning model to predict heart disease using an 1,025 patient record dataset. To achieve methodological rigor and generalization, 5- Fold Stratified Cross-Validation (CV) was used to evaluate all models, such as LightGBM and Random Forest. LightGBM model was stable and better in performance as it had Mean CV Accuracy of ±0.9620 ±0.0178. Integration of XAI (SHAP/LIME) is the means of creating clinical trust; analysis has confirmed maximum heart rate (thalach) and type of chest pain (cp) as medically significant characteristics. This framework supports the sustainable smart city healthcare through a highly transparent decision-support system, which manages the resources in optimizing scalable public health programs.

In 2022, the Korean government announced its intention to join the CPTPP, led by Japan and launched in 2018, which aims to achieve comprehensive tariff abolition and establish new trade norms. According to our analysis, as with traditional gravity model outcomes, it appears that the greater the per capita GDP and the shorter the distance between countries, the greater the effect on promoting trade. However, the trade-enhancing effect after the enforcement of the CPTPP has not yet been clearly confirmed. Nevertheless, if tariff abolition and reduction proceed as planned, the trade-enhancing effect of the CPTPP is expected to gradually expand. On the other hand, an increase in trade between Japan and CPTPP member countries may suggest a potential contraction in trade with non-member countries. This could pose a threat to Korean industries that need to establish a stable supply chain within the Asia-Pacific region. As Japan is South Korea's fourth largest trading partner, a detailed review of the impact of the CPTPP on future trade between Japan and its member countries should form the basis for discussions on Korea's possible participation in the CPTPP and the content of negotiations.

토양수분은 가뭄 발생과 해소를 매개하는 핵심 인자로서, 대기-지표-지하수로 이어지는 수문순환의 연결 고리 역할을 한다. 이처럼 가뭄 연구 및 분석을 위한 토양수분 자료 관측 센서를 설치하여 측정하고 있으나, 기상 및 통신 장애로 결측이 발생하여 자료 활용에 불편함을 겪고 있다. 본 연구에서는 해남·예산 지역에 설치된 토양수분 모니터링 시스템의 토양수분 결측 자료를 보간하기 위하여 먼저 강수 자료를 보간하고, 강수 특징변수로 머신러닝의 학습자료를 5개로 구축하여 학습 및 평가자료 정확도 결과를 비교·분석하였다. 연구 결과, 지연, 누적, 시계열 특징변수로 구성한 D, E 학습자료 기반 머신러닝 모형이 훈련·검증자료 정확도가 우수하였다. 평가자료 정확도는 D, E 학습자료 기반 XGB 모형이 우수하였으나, E 학습자료 기반 XGB 모형은 다른 조합 대비 더 많은 경우에서 우수한 정확도를 보였다. 따라서 E 학습자료 기반 XGB 모형을 활용하여 해남·예산 지역에 설치된 토양수분 모니터링 시스템의 10·20cm 깊이 토양수분 결측 자료를 보간하는게 적절하다고 판단하였다.

Soil moisture is a key variable governing drought onset and recovery and a critical link in the hydrological cycle connecting the atmosphere, land surface, and groundwater. Missing observations frequently occur in soil moisture monitoring systems due to meteorological and communication failures, limiting data usability. In this study, missing soil moisture data from monitoring systems in Haenam and Yesan were gap-filled by first correcting precipitation data and constructing five machine-learning training datasets using precipitation-based features. Model performance was evaluated using training, validation, and evaluation datasets. Results indicate that Training D and E datasets, incorporating lagged, accumulated, and time-series precipitation features, combined with the XGB algorithm, showed superior performance. The D– and E–XGB combinations also achieved high accuracy in the evaluation dataset, with the E–XGB model outperforming others in more cases. Therefore, the E-dataset XGB model is suitable for gap-filling 10- and 20-cm soil moisture data in the Haenam and Yesan regions.

This study investigates the use of machine learning techniques to estimate the value of studio apartments (Officetel), which are increasingly important as combined office and residential spaces for single-person households and freelancers. It aims to identify key variables affecting studio apartment prices and preprocess them for accurate predictions. Transaction data from studio apartments are used to compare the predictive performance of four methods: Multiple Regression Analysis, Random Forest, XG Boosting, and Deep Learning. The study seeks to determine the best-performing models for price estimation and aims to develop predictive models applicable to various real estate types and regions.

Although there have been studies using various algorithms on the delivery time prediction in the logistics business, those studies did not reflect various features such as region or product. In the case of delivery time prediction of a single model that does not reflect the features, the accuracy of delivery time prediction for a region with a high distribution is high, but the prediction accuracy is low for a region with a low distribution. To solve this problem, this paper proposes a method of classifying logistic patterns using clustering and selecting an optimal model for each logistic pattern. The proposed method consists of four steps. First, the derived variables such as reception day, delivery speed and delivery distance are created. Second, the data with the same pattern goes through clustering using K-means. Third, by comparing the performance of each model using six regression algorithms for each classified logistic pattern, an optimal model is selected and the model is stored. Lastly, the logistic pattern of the data to be predicted is classified and the optimal model stored for each pattern is loaded, and the result of delivery time prediction is provided through the model. Two experiments were performed to verify the proposed method. The e-commerce data from Brazil is used as verification data. From the experiment, the delivery time prediction error of the proposed model was smaller than that of the single regression model.

In this study, AAHT was estimated using machine learning algorithms of ensemble techniques to improve the limitations of existing studies. Among the machine learning algorithms, a random forest algorithm was used, and the traffic volume estimation accuracy was analyzed as MAPE 20.7%. It was analyzed that the higher the traffic volume level, the higher the traffic volume estimation accuracy, and the traffic volume level of 3,000 vehicles/hour or more was analyzed to be 8% or less of MAPE. It was analyzed that the accuracy secured at the current level was very high compared to the existing model.

Understanding an accurate trip demand by purpose is crucial for short-term regional planning but also long-term regional planning. In traditional approach, information of purpose-oriented trip demand has been derived from public survey in South Korea such as Travel Diary Survey. This type of data acquiring method maybe useful in a sense that it can capture a meaningful sample regarding to entire country, meanwhile it costs a tremendous amount of budget and time. In this research, we want to offer a novel framework for estimating purposeoriented trip demand with dynamic and effective fashions using data fusion in conjunction with Machine learning techniques. With primary results of this concept, this study showed how several state-of-the-art algorithms, including Deep neural network, UMAP, and random forest in conjunction with Genetic algorithm and Tabu-search for optimization, can contribute to this framework. Although tangible results are yet to come, we expect this framework can contribute to resilience planning such as COVID-19.

Predicting traffic accidents is a challenging task because taking into account uncertainty in modeling traffic accidents is not trivial. To address these issues, this article develops a hybrid modeling pipeline combining unsupervised and supervised learning to predict the level of hazardous road sites and explore the causality of accidents by controlling unobserved heterogeneity issues effectively. Traffic accident data for Won-ju province, Korea, from 2020 to 2021, and external factors affecting traffic accidents, such as average travel speed and weather information, are combined based on road links. Through the modeling pipeline, a clustering technique is adopted to capture unobserved heterogeneous information among roads. Since traffic accident data contains a wide variety of categorical and hierarchical features, ensemble methods such as boosting techniques were applied to handle heterogeneity issues among these features. To explore the relationship between the accident and determinant factors, are adopted to interpret the results of machine learning models. Model-agnostic methods, however, generally provide results based on images, this study also added a process that extracts texts from images to overcome compatible issues with existing road safety management systems.