Abstract: For the current attack path prediction method can not accurately reflect the impact of attack power on the subsequent path, an attack path prediction algorithm is proposed. First of all, the combination of the sort of attack path sample time series, re combination of attack path samples, using extreme learning machine classifier, and then use artificial fish fishing machine learning algorithm was optimized to improve the global search for network attack path samples and local optimization ability and avoid the local optimal solution. The simulation results show that the proposed method has better convergence, lower iteration number and higher prediction accuracy.