Research On Fiscal Revenue Forecast Based On Data Mining

Fiscal revenue is not only the most important form or source of revenue for the country,but also the government will submit a draft budget at the beginning of each year,according to the current year's fiscal revenue forecast,reasonable arrangement of all aspects of expenditure.So,how to predict,how to be more accurate current year or future fiscal revenue? This is exactly the question to be discussed in this paper.This paper is divided into two forecasting ideas,and selected a large population,located in the traffic fortress of Jinan City as an example,the above forecasting methods are demonstrated and popularized.(According to the data release unified requirements,the official according to the original Jinan City,the original Laiwu City two cities respectively issued).First of all,only use the historical data of fiscal revenue itself to forecast.(1)the GM(1,1)model is selected to predict the fiscal revenue of Jinan(former Jinan and former Laiwu)from2018 to 2021.results show that although the GM(1,1)model prediction accuracy grade is "good" and the posterior difference ratio is less than 0.35,which passed the model test.However,the relative error between the predicted value and the real value is generally around 10%.(2)In order to improve the accuracy,the BP neural network model is used to analyze the prediction.The results show that the relative error of BP neural network is about 4%,which is more accurate than that of GM(1,1)model.Secondly," Men are known by the company they keep ".If only through the financial revenue indicators to forecast,the results will inevitably be biased,so with the help of relevant factors to forecast it.(1)Select the main influencing factors: first adopt the Lasso model,in the initial selection of 9 indicators,screening.Although the Lasso model can solve the problem of multiple collinearity among the influencing factors,it tends to choose one of the characteristics,which leads to the instability of the results,and then selects the grey relational analysis method to screen out the main influencing factors of the financial revenue.The screening results are consistent with the qualitative analysis results,and the correlation degree between influencingfactors and financial revenue is given.Using influencing factors to forecast :6 indexes with correlation degree greater than 0.85 were selected by grey correlation analysis,and the combined model of GM(1,1)and BP neural network was used to forecast the fiscal revenue of Jinan from2018 to 2021.The prediction results are more accurate,and the relative error between the predicted value and the real value is less than 1%.Finally,“the beauty of beauty,beauty and beauty together”.The model and application of fiscal revenue forecast discussed in this paper can be extended to other fields,such as the forecast of enterprise profit,commodity price,infrastructure quantity and so on.Based on the BP neural network model,this paper discusses the expenditure in the key areas of Jinan people's livelihood.