基于 bp 神经网络和灰色模型的税务收入预测研究-计算机技术专业论文.docxVIP

摘 要 税收是以国家为主体，为实现其职能，凭借政治权力，按照法律规定，强制、 无偿取得财政收入的一种特定分配形式。随着经济的发展，对税收收入进行预测 显得越来越重要，它决定着税收计划的制定和经济的决策，然而由于税收成分及 相关因素的影响，对税收收入进行较为准确地预测，仍是一项需要深入研究的课 题。 本文首先介绍了常用预测模型的基本概念、方法，着重分析了自回归移动平 均模型、马尔可夫链模型和指数平滑法等预测中常用的方法和相关内容，并提出 了基于灰色算法模型的税务预测和基于BP神经网络的税务预测研究方法，分析了 其原理及算法。 在此基础上本文设计实现了预测程序，以陕西省商洛市商州区国税局的实际 税收数据为样本，进行仿真测试，对测试结果和实际数据进行比较，得出预测误 差达到精度标准要求的结果。最后对灰色模型和 BP 神经网络模型进行了对比，分 析了各自的优劣性及适用环境。 关键词：税收 预测 灰色模型 BP 神经网络 Abstract Tax is a kind of allotment relation of the public finance income forcibly and gratisly, which is carried out by a national political power according to the standard that national law enacts. With HYPERLINK /economic/ economic HYPERLINK /development/ development，it HYPERLINK /seem/ seem increasingly important in making prediction about tax revenue, which determines tax HYPERLINK /plan/ plans and HYPERLINK /economic/ Economic Decision. But because of the effects of tax classification and other related factors, it is one of subjects to be researched continuously to predict tax with great accuracy. Firstly HYPERLINK /firstly/ , we introduce HYPERLINK /introduce/ the basic conception and method HYPERLINK /method/ of commonly used prediction HYPERLINK /method/ models, with emphasis on the analysis HYPERLINK /analysis/ of the auto regressive moving average HYPERLINK /average/ prediction HYPERLINK /average/ model, Markov chain model HYPERLINK /model/ and Exponential Smoothing Method, and we propose a tax HYPERLINK /model/ revenue forecast based on grey model and related research method HYPERLINK /method/ based on BP neural HYPERLINK /method/ networks and analyses the principle and process. Bsaed on the above works, prediction program is designed and implemented in this paper, and we finish the HYPERLINK /simulation/ simulation testing according to the tax HYPERLINK /data/ data from Shangluo State Taxation Bureau, Shaanxi province. At last we carry on a contrast analysis of grey model and BP neural networks model and analyze their advantages,