Efficient neural network- and tree-based machine learning models for predicting shear capacity of RC slender walls

Slender reinforced concrete (RC) walls are popularly employed to improve the lateral loading capacity of high-rise buildings. Shear strength is an important target in designing RC walls subjected to horizontal loads such as earthquakes. This study aims to develop machine learning (ML) models for estimating the shear capacity of RC slender walls. A total of 154 test results, which were published in the literature, are gathered for training ML models. Two neural network-based models, i.e., Artificial neural network- Levenberg Marquardt (ANN-LM) and Artificial neural network-Bayesian regularization (ANN-BR), and two tree-based ML models including Random Forest (RF) and Gradient boosting regression tree (GBRT), are developed to predict the shear strength of RC walls. The predicted results obtained from ML models are then compared with those from empirical formulas in design codes. It shows that ML models are superior in predicting the shear capacity of RC slender walls compared to other code-based models, especially, RF and GBRT highly efficient models. Moreover, a practical graphical user interface tool is proposed to simplify the practical design of RC slender walls.