The GA neural network method based on the phase space reconstruction is proposed for electrical performance parameters prediction for the electromagnetic relays. This paper presents the phase space reconstruction method for the electrical performance parameter series of the electromagnetic relays by using the contanct resistance parameter and arc energy, and then builds the neural network prediction model of the electrical performance parameter series using genetic algorithm. The optimal prediction model parameters are given out by means of analyzing the parameter impacts on the prediction effects. The proposed method has the ability to discover the degradation process of electrical performances for the electromagentic relay in the high dimensional phase space and obtains the better precision of prediction. The experimental results show that the proposed method is feasible and effective.