Committee integration of optimized models for prediction of compressive strength of masonry structures

Abstract An accurate computation of the compressive strength of masonry structures (CSMS) is an overarching factor in design and construction of masonry structures (MS). This considerable significance compels researchers to propose an appropriate, reliable and, more generalized method whereby the precise value of CSMS is calculated. In the current study, a committee machine (CM) with optimized elements is constructed, thereby extracting a non-linear relationship between CSMS with compressive strength (CS) of mortar and brick. In order to accomplish this objective, three intelligent models viz. neural network (NN), fuzzy inference system (FIS), and support vector regression (SVR) are firstly optimized with bat-inspired algorithm (BA), and these improved models are subsequently applied for estimation of CSMS. BA is hybridized with intelligent models for extracting the best values of weights and biases of NN, membership’s functions of FIS, and user-defined parameters of SVR. Then, CM is utilized for amalgamating the outputs of three optimized models (OMs) incl. optimized neural network (ONN), optimized fuzzy inference system (OFIS), and optimized support vector regression (OSVR). BA is also embedded in the structure of CM, thereby determining the optimal contribution of each optimized model in the final prediction. Data sets including 96 records accessible in the literature are used to learn and evaluate the constructed models. Appraisal of the accuracy based on statistical parameters verified that the CM could effectively improve the prediction accuracy of the OMs and also has a better performance compared to commonly well-known predictive correlations. This study also proved that CM with optimized elements is a very convenient approach for mapping nonlinear functions between CSMS and CS of brick and mortar.