Artificial neural networks (ANN) are known to be increasingly popular and used in several engineering applications, such as in the civil engineering field. In this study, this method was used to develop an optimal model to predict the shear strength of concrete using the experimental data sets. All the data sets were trained and tested using ANN to obtain the prediction of the shear strength of concrete material. The model ANN was trained and tested using test data sets obtained from 51 concrete mixes from previous experimental data sets. 33 (65%) concrete mixes data sets were chosen randomly and used as input for training. The remaining 18 (35%) mixes data were divided equally into testing and validation data sets. Feed-forward backpropagation was chosen for the neural network design and LevenbergMarquardt was used as the learning algorithm. An S-shaped sigmoid function was used to predict the probability as output between the range 0 to 1. Ten different types of architecture networks with different types of structures and neurons number were used to obtain the best model. The optimal ANN architecture (33-10-1) was found to have the highest correlation coefficient (R) of 0.99888 and the lowest mean square error (MSE) 0.00085. The shear strength based on the ANN model perfectly matched the values of the experimental data sets.