Enhancing failure mode classification of RC beam-column joints using logistic regression and hybrid sampling strategy

Abstract

Seismic failure mode of reinforced concrete (RC) beam-column joints (BCJs) is crucial for the safety and integrity of RC building or structure withstanding seismic forces. However, traditional classification methods are biased towards estimating majority samples and often misclassify minority samples due to imbalanced data distributions, leading to unexpected classifications for seismic failure modes of BCJs. To address the challenge of imbalanced data in classifying seismic failure modes of BCJs, an innovative imbalanced classification method based on logistic regression (LR) and hybrid sampling strategy is proposed. The method was compared with traditional shear-resistance design methods and LR models based on 197 sets of experimental data. Results demonstrate that the proposed method consistently outperforms traditional approaches. Specifically, the proposed method maintains higher values for K_a and M_cc, even as the class imbalance ratio increases, indicating its robustness in handling imbalanced data. The proposed imbalanced classification method offers several advantages over traditional approaches and a promising tool for accurately classifying seismic failure modes of BCJs, even in the presence of imbalanced data.