PNN神经网络评价方法本科设计说明

摘 要企业评价是检验企业经营效果、校正进展方向的重要手段，对企业的成败具有决定意义。如何用定量的方法准确的评价企业是人们一直在讨论的问题，对企业的正确评价可以指导我们的投资决策。概率神经网络是一种可用于模式分类的神经网络，本文将利用概率神经网络对企业进行评价，将企业分为“好”和“差”两类。本文综述了用于企业评价的方法、概率神经网络原理与结构，并将概率神经网络应用到对企业的评价中。首先，利用逐步回归方法从 60 家上市公司的 12 个财务指标进行筛选得到与企业收益率显著相关的 4 个指标。其次，以这 4 个财务指标为输入向量，以企业的好差情况为输出向量，利用 60 家上市数据建立并训练 PNN 神经网络。最后，对训练样本和测试样本进行了仿真，仿真结果表明：利用概率神经网络进行企业分类评价的是一种非常有效的方法。在结论中，对本文容进行了总结,并对今后的工作做了更进一步的展望。关键词：评价方法， PNN 神经网络，SPSS，MATLABABSTRACTEnterprise’s evaluation is animportant means to inspect the business’s results and checkthe developmental direction . How to use quantitative method to evaluate enterprises accurately is a question which people have been researched all along, and thecorrect evaluation of enterprises can guide our investment decisions. Probabilistic neural network is a net which can be used for the classification.In this paper, the probabilistic neural network is used to evaluate enterprises,and enterprisesare divided into "the good" and "the bad" . This paper summarizes the methods which areapplied to enterprise’s evaluation, andintroduces the probabilistic neural network principle and structure, then probabilistic neural network is applied to evaluation of enterprises. First of all, four amongtwelve financial indicators ofthe 60 listed companies are significantly correlated with the yields on corporate bondsby using stepwise regression method. Secondly,four financial indicators are regarded as the input vector and the cases of ...