In this paper, discrete element modelling and artificial neural networks (ANNs) were adopted to evaluate seismic performance of a masonry-infilled reinforced concrete (RC) frame. Discrete element models were developed to simulate quasi-static tests of infilled frames and were validated by using experimental data from the literature in terms of ultimate bearing capacities, initial stiffnesses and hysteresis curves. Parametric analysis was conducted based on the validated models to further investigate the influence factors of infilled RC frames and to collect a database for training ANN models. Subsequently, 440 datasets were divided into a training set (70 %), a validating set (15 %), and a testing set (15 %). Back propagation neural network (BPNN) and radial basis function neural network (RBFNN) were employed to predict the seismic behavior of a masonry-infilled RC frame. The BPNN model used the Levenberg-Marquardt algorithm exhibited the highest coefficient of determination, and it was better than the RBFNN model with more neurons. Eventually, a practical ANN model was proposed to evaluate the seismic performance of a masonry-infilled RC frame.