In this paper, we investigate predicting cascading failure propagation path in complex networks by using neural networks. A cascading failure simulation model considering the cumulative effect of overloading of components over time is firstly established. Then, a cascading failure prediction model combining an attention mechanism and long and short-term memory (LSTM) neural networks is proposed. We generate cascading failure cases as the ground truth with the cascading failure simulation model and make predictions with the Attention-LSTM neural network in synthesized complex networks. Simulation results show that our proposed method can predict the path of cascading failure propagation quickly and accurately.