大多分类问题由于各特征之间存在交互性等不确定性问题,使一般的神经网络分类算法难以取得好的分类效果。该文通过构造新特征,优化了神经网络的输入,并给出一个优化的基于神经网络的分类算法。文章还引入了分类问题的离散度的概念。并结合实例进行分析。 Most classification problems due to the interaction between the various characteristics of the uncertainty, so that the general classification algorithm for neural network is difficult to achieve good classification results. By constructing new features, this paper optimizes the input of neural network and gives an optimized classification algorithm based on neural network. The article also introduces the concept of dispersion of classification problems. And combined with examples for analysis.