In this paper, machine learning models were applied to predict the seismic response of steel frame structures. Both geometric and material nonlinearities were considered in the structural analysis, and nonlinear inelastic dynamic analysis was performed. The ground acceleration response of the El Centro earthquake was applied to obtain the displacement of the top floor, which was used as the dataset for the machine learning methods. Learning was performed using two methods: Decision Tree and Random Forest, and their efficiency was demonstrated through application to 2-story and 6-story 3-D steel frame structure examples.

본 연구에서는 2 016년부터 2 02 0년까지 내륙 관측소 중 안개 최다발 지역인 안동을 대상으로 XGBoost-DART 머신러닝 알고리즘을 이용하여 1 시간 후 안개 유무를 예측하였다. 기상자료, 농업관측자료, 추가 파생자료와 각 자료 를 오버 샘플링한 확장자료, 총 6개의 데이터 세트를 사용하였다. 목측으로 획득한 기상현상번호와 시정계 관측으로 측 정된 시정거리 자료를 각각 안개 유[1]무[0]로 이진 범주화하였다. 총 12개의 머신러닝 모델링 실험을 설계하였고, 안개 가 사회와 지역사회에 미치는 유해성을 고려하여 모델의 성능은 재현율과 AUC-ROC를 중심으로 평가하였다. 전체적으 로, 오버샘플링한 기상자료와 기상현상번호 기반의 예측 목표를 조합한 실험이 최고 성능을 보였다. 이 연구 결과는 머 신러닝 알고리즘을 활용한 안개 예측에 있어서, 목측으로 획득한 기상현상번호의 중요성을 암시한다.

PURPOSES : In this study, model-agnostic methods are applied for interpreting machine learning models, such as the feature global effect, the importance of a feature, the joint effects of features, and explaining individual predictions. METHODS : Model-agnostic global interpretation techniques, such as partial dependence plot (PDP), accumulated local effect (ALE), feature interaction (H-statistics), and permutation feature importance, were applied to describe the average behavior of a machine learning model. Moreover, local model-agnostic interpretation methods, individual conditional expectation curves (ICE), local surrogate models (LIME), and Shapley values were used to explain individual predictions. RESULTS : As global interpretations, PDP and ALE-Plot demonstrated the relationship between a feature and the prediction of a machine learning model, where the feature interaction estimated whether one feature depended on the other feature, and the permutation feature importance measured the importance of a feature. For local interpretations, ICE exhibited how changing a feature changes the interested instance’s prediction, LIME explained the relationship between a feature and the instance’s prediction by replacing the machine model with a locally interpretable model, and Shapley values presented how to fairly contribute to the instance’s prediction among the features. CONCLUSIONS : Model-agnostic methods contribute to understanding the general relationship between features and a prediction or debut a model from the global and/or local perspective, securing the reliability of the learning model.