본 연구에서는 2011년부터 2024년까지 새만금 지역의 4개 주요 지점(만경강, 동진강, 신시갑문, 가력갑문)에서 수집된 수질 자료 를 이용하여 용존산소(DO)와 총유기탄소(TOC) 예측을 위한 XGBoost 기반 모델을 구축하고, SHAP 분석을 통해 변수별 상대적 설명력을 평가하였다. 모델은 DO에서 R² 0.89–0.95, TOC에서 0.88–0.95의 높은 예측 성능과 낮은 평균제곱오차(MSE)를 보여, 예측의 신뢰성을 확 인하였다. SHAP 분석 결과, 하천 지점에서는 pH와 수온이 DO 예측에서 가장 높은 설명력을 보였고, 염분의 영향은 미미하여 외해수 유입 이 제한적임을 나타냈다. 반면, 기수 지점에서는 수온이 DO 예측의 주요 요인으로, 염분은 보조 요인으로 작용하였다. 연도별 분석에서는 하천 지점에서 pH의 기여도가 2016년 이후 감소하고 수온의 비중이 2019년 이후 다소 높아지는 경향이 일부 나타났으나, 전체적으로는 명 확한 증가·감소 추세가 확인되지 않았다. TOC 예측에서는 하천 지점에서 COD와 chlorophyll-a가, 기수 지점에서는 chlorophyll-a와 염분이 상 대적으로 중요한 설명 변수로 확인되었다. COD의 기여도는 2017–2018년에 낮았다가 2019–2021년에 높아진 후 최근 다시 감소하는 등 일부 연도에서 변동이 나타났으나, 뚜렷한 장기 경향은 관찰되지 않았다. 이러한 결과는 연도별로 변수의 상대적 설명력에 세부적인 변동 은 존재하지만 전체적으로 일관된 추세는 아직 확립되지 않았음을 보여준다. 이처럼 연도별 변동성과 불확실성이 공존하는 환경에서, XGBoost와 SHAP을 결합한 접근법은 각 변수의 상대적 중요도와 시기별 변화를 정량적으로 평가할 수 있는 유용한 분석 틀을 제공한다.

Using highway accident data, this study predicts the probability of rollover, overturning, and fire accidents and identifies the related risk factors. Whereas existing studies rely primarily on limited explanatory variables and classical statistical models, this study simultaneously enhances predictive performance and interpretability by applying and comparing machine learning-based nonlinear prediction-analysis systems (XGBoost and Shapley additive explanations) with logistic regression, which offers advantages in statistical reasoning. The analysis identifies speeding, segment characteristics (tunnel, ramp, shoulder), and vehicle type (SUV, truck, trailer, and tank lorry) as common key risk factors. These results suggest the necessity of establishing a multilayered management system for speeding, improving facilities centered on high-risk sections (tunnel in/out, ramp, and downhill), performing custom inspections for each vehicle type (load, tire, and brake system), and improving driving behavior (enhancing forward attention, introducing a drowsiness warning system, etc.). This study provides a datadriven empirical basis for identifying the causes of major highway accidents and for designing effective prevention policies.

In this study, we aim to classify personal mobility (PM)-related traffic crash data into four categories: PM-to-vehicle, PM-to-pedestrian, PM-single, and vehicle-to-PM crashes, and analyze the factors influencing the severity of each crash type. To overcome the limitations of existing studies in explaining the impact of independent variables on ordinal dependent variables, a random forest model was combined with the Shapley additive explanation technique. This approach visualizes the influence of independent variables on a dependent variable, providing clearer insights and enhancing interpretability. The analysis of PM traffic accidents, categorized into at-fault, single-vehicle, and victim accidents, revealed distinct key factors for each type. The main contributors to the severity of crashes caused by PM are traffic violations by teenagers and collisions with elderly pedestrians. Single-vehicle accidents were predominantly caused by overturn incidents, with inadequate driving skills among PM users aged 40 years and older, and significantly increasing severity. Victim accidents primarily occur at intersections, where the behavior of the at-fault driver and age of the PM user are critical factors influencing the severity. We identified various factors influencing the severity of PM crashes by type, highlighting the need for tailored policy measures. Proposed policies include physically separating bicycle–pedestrian shared spaces and strictly regulating illegal PM sidewalk riding, introducing PM licenses for teenagers to ensure compliance with traffic rules, and implementing regular safety education programs for all age groups. Although this study applied a new analytical technique, it relied on limited crash data, thus limiting the results to estimates.

Recently, the development of computer vision with deep learning has made object detection using images applicable to diverse fields, such as medical care, manufacturing, and transportation. The manufacturing industry is saving time and money by applying computer vision technology to detect defects or issues that may occur during the manufacturing and inspection process. Annotations of collected images and their location information are required for computer vision technology. However, manually labeling large amounts of images is time-consuming, expensive, and can vary among workers, which may affect annotation quality and cause inaccurate performance. This paper proposes a process that can automatically collect annotations and location information for images using eXplainable AI, without manual annotation. If applied to the manufacturing industry, this process is thought to save the time and cost required for image annotation collection and collect relatively high-quality annotation information.

Recently, the importance of preventive maintenance has been emerging since failures in a complex system are automatically detected due to the development of artificial intelligence techniques and sensor technology. Therefore, prognostic and health management (PHM) is being actively studied, and prediction of the remaining useful life (RUL) of the system is being one of the most important tasks. A lot of researches has been conducted to predict the RUL. Deep learning models have been developed to improve prediction performance, but studies on identifying the importance of features are not carried out. It is very meaningful to extract and interpret features that affect failures while improving the predictive accuracy of RUL is important. In this paper, a total of six popular deep learning models were employed to predict the RUL, and identified important variables for each model through SHAP (Shapley Additive explanations) that one of the explainable artificial intelligence (XAI). Moreover, the fluctuations and trends of prediction performance according to the number of variables were identified. This paper can suggest the possibility of explainability of various deep learning models, and the application of XAI can be demonstrated. Also, through this proposed method, it is expected that the possibility of utilizing SHAP as a feature selection method.