Sandwich honeycomb blast wall with the superior energy absorption characteristics could be applied to anti-blast protection of offshore oil and gas platform. Recently, machine learning technique, as an alternative to traditional Finite Element Analysis (FEA), is becoming a powerful tool in understanding structural performance. The aim of this study is to propose a novel data-driven model for fast predictions of structure response of sandwich blast wall under explosion, including the maximum deformation δmax, the maximum internal energy Emax and the energy absorption ratio RE. FEA was firstly conducted to generate sample data for the proposed model by using LS-DYNA software. An improved artificial electric field algorithm (IAEFA) was combined with artificial neural network (ANN) to develop an IAEFA-ANN model. In case studies, the prediction results of IAEFA-ANN approximate the results of FEA well, and comparison results of 9 different ANN models demonstrate that IAEFA-ANN model has the best performance. Then, the effects of explosion parameters (including explosive charge mass Me, stand-off distance d and concave angles θ) on structural response prediction results were investigated by using the proposed model. Moreover, the IAEFA-ANN model was applied to conducting optimal design of the blast wall to obtain optimum anti-blast performance.