Structural failure classification for the reinforced concrete (RC) buildings is one of the machine learning challenging tasks. Several successful studies were conducted to train the Neural Network (NN) with well-known optimization techniques. In the current work, a cuckoo search (CS) based classification model of structural failure of the RC buildings was proposed. The proposed NN-CS system was compared to well-known models, namely the Multilayer perceptron feed-forward network (MLP-FFN) trained with scaled conjugate gradient descent and the NN supported by the Particle swarm optimization algorithm (NN-PSO). The performance metrics, including the accuracy, precision, recall, and F-measure were calculated. The experimental results established the superiority of the proposed NN-CS with reasonable improvement (93.33% accuracy) compared to the other models. © 2019 The authors and IOS Press. All rights reserved.