Abstract: In this paper,we propose a new prediction method of chaotic frequency hopping codes based on the optimized echo state network,in order to overcome the problem of low recognition accuracy and vast computation of mathematic prediction models with their incapability of memorization.This method inherits the excellent performance of ESN and the improved genetic algorithm solves the problem of dynamic reservoir parameters at the same time,making the mentod has more pertinence and better prediction performance.With optimized network parameters and sample data chosen from Logistic-kent mapping,Lorenz system and Mackey-Glass system frequency hopping codes,simulation experiments have been conducted. The experiments proves the validity of the proposed method In the comparison with other literatures.