神经元网络由于分类精度高及具有并行性、容错性等优点，在模式识别、组合优化等许多领域得到了成功的应用，但是由于它的结果难以理解而影响了它的可信度及其实用性。该文主要是应用多策略学习方法，经过对神经网络进行训练、剪枝，然后从中抽取出符号规则，以解决神经网络的这种“黑箱”问题。文中讨论了几种方法的紧密结合，有效地应用了各种方法的优点，实验也证明了方法的有效性。 Due to its high classification accuracy, parallelism and fault tolerance, neuronal networks have been successfully applied in many fields, such as pattern recognition and combinatorial optimization, but their credibility and utility have been affected because of its incomprehensible results Sex. This paper mainly applies multi-strategy learning method, after training and pruning the neural network, and then extract the symbol rules from it to solve this “black box” problem of neural network. The paper discusses the close combination of several methods and effectively applies the advantages of various methods. The experiment also proves the effectiveness of the method.