Abstract
Rectangular reinforced concrete (RC) shear walls are crucial for seismic resistance in high-rise buildings. Characteristic points on the skeleton curve from pseudo-static tests are key for evaluating seismic performance. Traditional models struggle with the high-dimensional, nonlinear relationships between these points and component dimensions. An interpretable empirical guidance machine learning (IEG-ML) was proposed for predicting these feature points. IEG-ML aligns with empirical trends from extensive experiments and is explainable via explicit formulas. Trained on a self-built dataset of 184 samples, IEG-ML accuracy and efficiency are enhanced using a population optimization algorithm. The model identifies the importance of feature points and component factors, providing a dominant explicable formula. Results show IEG-ML high accuracy and efficiency, particularly with a backpropagation network optimized by the dung beetle algorithm (DBO), making it a robust tool for seismic evaluation.