Research On Data Mining Algorithm For Tax Forecasting

With the development of economy and information technology,the application of data mining technology is more and more widely in various fields.As an important component of the economy and society,tax plays a very important role in the national fiscal revenue.At present,many scholars have applied the data mining technology to study the problems of tax forecasting,tax collection and management and decision making.The development of the work budget and the formulation of the tax plan should be based on the tax forecasting results,so the tax forecasting algorithm has been highly valued by the researchers worked in the tax industry.It is urgent to solve the problem of how to make good use of the data with small sample and multi dimension in the application of the tax.In order to improve the accuracy of tax forecast,this paper puts forward a method of using data mining technology to forecast the tax revenue.First of all,in order to obtain the high quality tax index sample set,the historical tax data need to be preprocessed.The correlation analysis and stepwise regression analysis were used to reduce the dimension of the tax index,and to solve the key indicators of different tax categories.Under the condition of the default parameters of support vector machine algorithm,prediction algorithm have the problem of poor accuracy.Aiming at the above issues,this paper proposes the means of using grid search method to improve the parameters of support vector machine algorithm,so as to solve revenue optimization prediction algorithm based on support vector machine.Secondly,aiming at the prediction and predictability problem of forecasting algorithm,this paper puts forward a kind of optimal combination forecasting algorithm.Firstly,The GM(1,1)forecasting algorithm based on grey series and the ARIMA tax forecasting algorithm based on time series are established respectively,and then obtain the tax forecasting values of the two algorithms,the above prediction results and the real tax data sets compose the combination algorithm's experimental data set.Then,the multi-level grid search method is used to solve the weight coefficient which can minimize the prediction error,and then achieve the optimal combination forecasting algorithm.Finally,the real tax data are used to verify the optimal support vector machine algorithm and the optimal combination forecasting algorithm.The simulation results show that the prediction accuracy of the two algorithms is higher than the unmodified ones,and the validity of the algorithm is verified.