创业板研发投入的神经网络预测：是采用RBF模型还是BP模型？

The research on BP and RBF neural network prediction model for R D of GEM

摘要:

创业板公司成立时间较短,企业规模较小,其研发投入的影响因素较多,本文利用营业净利率、每股收益、董监高年薪、可持续增长率、资产负债率、现金流量净额和GDP这些指标,运用径向基神经网络(RBF)和逆传播神经网络(BP)方法,构建了一个训练完成的神经网络模型,研究发现RBF神经网络模型比BP神经网络模型具有更好的拟合、预测效果。

Abstract:

GEM enterprises are mostly setting up in a short time,having small business size and their performance is not outstanding,but most of them have a high growth. Their high growth performance are reflected in the ability of research and development on sensitive market reaction .However,there are many factors affecting R D investment. Based on the annual salary of the directors,supervisors and senior,sustainable growth,rate of assets and liabilities,net cash flow and GDP,using BP and RBF neural network model,we construct a neural network model which is training completed. It is found that the RBF neural network has better fitting and forecasting effect than BP neural network.