基于BP网络的GDP预测数据预处理方法研究

摘要

Data prerocessing is one of the critical processes during modeling. Different methods of data preprocessing will lead to different results in the same model. The article discusses six methods and two new methods of data preprocessing in GDP forcasting model based on BP NN using MATLABTCX)L. After comparing and analyzing the results, the method of cut-division is a simple and most precise one.%数据预处理是决定建模成功与否的关键步骤之一,而同一个模型用不同的预处理方法所得到的结果也会不一样.文章基于BP神经网络的GDP预测模型,探讨了已有的6种预处理方法,在这些方法之上又新增了2种不同的方法,最后用这些不同方法在MATLAB下实现,将所得结果作比较与分析,文章提出的减比法是一种既简单准确度又高的方法.