本文提出一种基于最优权值的选择性神经网络集成构造方法,在训练出个体神经网络之后,使用遗传算法计算出这些网络在加权平均方法中对应的最优权值,然后选择权值大于一定阈值的部分网络使用简单平均方法组成神经网络集成,理论分析和实验结果表明,与传统方法相比,本文方法使用部分网络能够取得更好的效果。 In this paper, we propose a method of constructing a selective neural network based on the optimal weight. After training the neural network, we use the genetic algorithm to calculate the corresponding optimal weights of these networks in the weighted average method. Then, Some networks with a certain threshold value use a simple and average method to compose neural network integration. Theoretical analysis and experimental results show that this method can achieve better results than some traditional networks.