Research And Application Of Prediction And Analysis Method For Fiscal And Tax Categories

With the development of computer and Internet technology,we have entered an unprecedented era of big data.Big data across all walks of life,the financial and tax industry is also living in it.Inside the big data must bury huge data secret,how to excavate,discover and use the value of these data is a big challenge that we face.Tax revenue is an important source of fiscal revenue in China,and the prediction of tax revenue is also an important basis for adjusting the direction of economic development and formulating the development policy.It can also provide guarantee for the government at all levels to grasp the financial operation and formulate an accurate and scientific tax budget.China's tax system is huge and complex,different taxes and different economic indicators there is a deep level of inevitable connection,so tax forecast need to different categories of taxes for personalized prediction analysis.This paper selects several taxes which account for a large proportion of tax revenue,and carries on the forecast analysis separately,including: the business tax,the enterprise income tax and the individual income tax.Including economic indicators related to these taxes: GDP(GDP),added value,fixed asset investment,industrial investment,investment in real estate development,import and export volume,above scale industrial added value of third industries(the main business income of 20 million or more enterprises),consumption,total retail sales of social consumer product sales of real estate development area,the consumer price index for industrial use.This thesis uses data mining tools,using linear regression,neural network and time series and other algorithms to predict.Multiple linear regression model is the relationship between tax revenue and economic indicators,and to analyze the impact of economic indicators on the tax of importance,and ultimately the formation of multiple regression equation,using the regression equation for the next phase of the direct tax forecasting.Neural network and time series can not output model,use data mining tools to model training,and output prediction results,and then input the prediction results to the application platform.Based on the historical data of Beijing city tax as the data set for model training.Application platform features include two models,one is the application of regression model,through the model training to determine the regression equation coefficients and constants and input to the platform,through the platform only need to enter the next phase of the economic index data,the platform can automatically calculate the next phase of the tax revenue,greatly improve the prediction efficiency.The two is to predict data management,mainly to predict the results of each model directly input to the application platform,and then through the platform to display,this kind of function can be applied to all models.Through this platform,we can realize the rapid tax forecast,and at the same time,the functional tax categories can be expanded,which can increase other new tax forecast,such as real estate tax,deed tax and so on.