Prediction And Analysis Of Fiscal Revenue In Shandong Province Based On Grey Markov-BP Neural Network Combination Model

Local fiscal revenue comprehensively reflects the level of local economic development and is an important source of funds for the government to build infrastructure.Although the economic performance of Shandong Province is stable and improving,tax breaks and fee cuts and the impact of the COVID-19 epidemic have slowed fiscal revenue growth and the gap between fiscal revenues and public spending is becoming more pronounced.The establishment of fiscal revenue predicting mechanism and scientific predict of fiscal revenue will help the government maintain a reasonable balance of revenue and expenditure and promote the development of social economy.This paper takes the local fiscal revenue and influencing factors of Shandong Province from 1994 to 2021 as the research object,combines qualitative and quantitative research methods to select variables for the influencing factors.The combination model is further established to predict the fiscal revenue of Shandong Province in 2022 and 2023.The results show that the combination model’s prediction accuracy surpasses that of the single model.The research content is as follows:In this paper,17 influencing factors of fiscal revenue are selected.The qualitative index—science and technology innovation policy is innovatively added,and the Grey Variable Weight Clustering Method is used to quantify it.Descriptive analysis and correlation analysis are carried out on all data,and the characteristics of the data are preliminarily summarized.In order to reduce modeling errors,three methods,namely Lasso,Adaptive Lasso and Random Forest,are used to select influencing factors.The variables selected by the three methods are respectively used to establish the Lasso-BP,Adaptive Lasso-BP and RF-BP models for fiscal revenue fitting.The results show that the R square of the RF-BP model is maximum and the Mean Absolute Error is minimum.Therefore,the selecting results of Random Forest are used in the subsequent prediction model.This paper sets up a single model and a combination model to predict fiscal revenue.On the one hand,the single Grey predicting model is used to fit the fiscal revenue directly and get the predicted value.On the other hand,combined with Support Vector Regression(SVR)and BP Neural Network,the Grey Markov-SVR and Grey Markov-BP Neural Network combination prediction models are established according to the influencing factors to predict the fiscal revenue of Shandong Province in 2022 and 2023.Compared to the single model,the error of the combination model is less and the prediction accuracy is higher.Moreover,BP Neural Network is superior to SVR in prediction accuracy and error,and has higher reference value.Based on the above analysis results,this paper combines with the economic situation of Shandong Province to put forward the corresponding development suggestions.The scientific predicting model provides reference for the local government to formulate relevant policies,and is benefit to promote the economic development of society.