Computational approach towards shear strength prediction of squat RC walls implementing ensemble and hybrid SVR paradigms

Squat-reinforced concrete (RC) walls are shear walls with a low aspect ratio and are vital structural components for nuclear structures and conventional buildings owing to their considerable strength in resisting lateral seismic load. Existing empirical equations and design codes carry substantial discrepancies in terms of accuracy in estimating the shear strength (V_n) of squat RC walls. This work employs a hybrid paradigm of support vector regressor (SVR) and Harris hawk optimisation (HHO) algorithm, i.e., SVR-HHO, to predict the V_n of squat RC walls. The outcomes of the SVR-HHO framework were compared with some of the widely used soft computing paradigms, namely the generalised linear model, decision tree, random forest, gradient boosting tree, and standard SVR. Overall outcomes indicate that the developed SVR-HHO paradigm attained the most precise estimation of V_n of squat RC walls with 98 % (in terms of R-index) accuracy in the testing phase. Moreover, the SVR-HHO framework was validated based on parametric and sensitivity analyses. The outcomes were also compared with the present codes and empirical models previously available in the literature. Compared to the developed models, SVR-HHO demonstrated robust predictions. The minimum coefficient of variation (COV in %) was observed to be 23. 02 % for SVR-HHO, followed by ACSE (41.2 %) and ACI (45.3 %). To facilitate a quick estimation of the results, the developed model is also attached as supplementary materials.