Earthquakes are natural disasters which may result in heavy losses. Accurate prediction of the time and intensity of future earthquakes can lead to inimizing losses due to earthquakes. A number of earthquake predictions have been proposed based on mathematical and statistical models. In this paper, we present an earthquake prediction technique using Bat Algorithm (BA) and Feed Forward Neural Net- work (FFNN). The BA is used to train the weights of the
FFNN to predict future earthquakes on the basis of past input data. Experimental results show that our proposed approach is highly comparable and more stable than Back Propagation Neural Network (BPNN) with respect to accuracy.