JSA-LSSVR analysis on volume expansion of cement paste with fly ash and MgO expansive additive

There is little study on using machine-learning methods to estimate cement paste volume expansion (V_e) with fly ash (FA) and MgO expansive addition (MEA). The goal of this work was to create and validate machine learning methods for the evaluation of V_e by using a collection of data that included 170 experimental findings that were taken from published studies. The least-square support vector regression (LSSVR) was developed to accomplish this goal. The efficacy of LSSVR is significantly impacted by its hyperparameters, determined by the flow direction algorithm (FDA) and the jellyfish search algorithm (JSA) (LSF and LSJ). The values of this analysis depicted that Age attribute has the highest value of feature importance at 0.917 compared to others, where the MEA, FA, and PC are in the following ranks by obtaining 0.868, 0.827, and 0.795, respectively. The findings suggest that the LSF and LSJ approaches possess substantial promise for reliably predicting the V_e. The value of error-based metrics indicators for LSJ is considerably lighter compared to the LSF, by receiving almost 50% variance in the training section and roughly 150% in the testing section.