Study On The Tax Forecast Based On BP Neural Network

Tax revenue is not only the means of GDP distribution and macro-control but also could be the financial fund that government raises in order to provide the public products and public service, and it can adjust the profit distribution among different economies. Tax grew speedily with 17.13% per year after the reform of tax distribution from 1994, and it is always the leading role of total financial revenue. Since the establishment of market economic system,tax reform and making tax plans are always the focus point on government’s economic work,and that expresses more apparently during the reform period of country economic transition at present. Tax is a complex product in the process of economic development, so it is very important to analyze data accurately for the tax department decision. However, appropriate tax plan is based on way of tax forecast. Thus, we must abandon the traditional base-plus-growth method to set up tax plan, and need form a logistic, efficient and easy-operated advanced tax forecast system. Strong non-line fitting ability of BP Neural network is a definitely advantage at the tax forecast, and will be the main tool that government conducts on tax forecast.Firstly, this article determines tax revenue forecast target system after learning the concepts about tax forecast and tax plan. Secondly, the tax forecast model is structured upon BP Neural network which is based on expounding BP Neural network basic theory,forecasting fundamental and steps, feasibility. Thirdly, the article is with the help of computer codes at MatlabR2010 b to conduct network training and emulation, analyze and expound the error of forecasting value and target value from the viewing angle of empirical evidence, in which the sample data is the tax figures during 1995-2014.At last, the article uses the forecast model to realize the whole country tax revenue and local tax of 31 provincial administrative regions forecast in three years. It turns out to be reasonable one more time that BP Neural network can forecast tax revenue more accurately.Moreover, it also can offer reasonable suggestions for launching the policy that replacing the business tax with value-added tax to forecast the tax revenue in the 13 th Five-Year Plan.