论文部分内容阅读
论文针对经济预测通常表现为复杂的非线性这种特性,提出了一种基于自组织过程神经元网络(FPNN)和改进的BP神经网络建立的经济预测模型方法。自组织过程神经元网络(FPNN)由输入层、竞争层和输出层组成。FPNN筛选出对因变量(网络输出)最有影响作用的变量(自变量)之后作为改进的BP算法网络的输入节点,再用进行学习。该模型不仅克服了时间序列预测模型只能进行线性预测的不足,而且还避免了传统神经网络的固有缺陷。以2001年到2004年国内生产总值作为预测分析样本,并对预测结果和实际值进行了比较分析,结果验证了该方法的有效性。
In this paper, aiming at the characteristic that economic forecast usually shows complicated nonlinearity, this paper proposes an economic forecasting model based on self-organizing process neural network (FPNN) and improved BP neural network. The self-organizing process neural network (FPNN) consists of input layer, competing layer and output layer. FPNN screened variables (network output) the most influential variables (independent variables) as an improved BP algorithm network input node, and then learning. This model not only overcomes the shortcomings that the time series prediction model can only perform linear prediction, but also avoids the inherent defects of the traditional neural network. Taking the GDP from 2001 to 2004 as the predictive analytic sample, the forecast results and the actual values were compared and analyzed. The results verify the effectiveness of the method.