Revenue Forecasts Based On Neural Network Research

There is no doubt that accurate data analysis is very important in terms of the tax department. Data analysis by policy makers at all levels, the next step in the work of tax authorities to make a more comprehensive analysis and the most sensible decision-making, but the true analysis of survey data analysis is not the only step in the process. To get the most accurate analysis and it reports to decision-makers, you must be able to design data analysis project, data collection, to easily access and manage data, analyze data, and then share your analysis. The analysis process should include more than these steps.In order to complete an end-to-analysis, tax analysis need to set will not hinder their work, or reduce their productivity tools. But, in fact, many analysts are able to give them the lack of a broad, flexible methods of analysis solution, even harder to have a coverage analysis of each step in the process of solution. As a result, a new research paper from the perspective of the establishment of Jilin Province, two tax information than the data warehouse built on this basis, BP neural network model to forecast tax revenues.First of all,I am discussing towards to the background,purposes of the research and the meaning,briefly introduced the neural network technology in predicting the field of application, research status at home and abroad a certain amount of research and pointed to my main tasks.Secondly, the paper introduces the basic principles of data mining, as well as the basics of BP neural networks and basic network model and its learning process.Third, the thesis of the data warehouse development objectives, application framework, system architecture, design methods, and detailed the implementation process were introduced. Then, the paper discusses in detail the tax based on BP neural network prediction model of the building. Mainly include:Based on BP neural network prediction model for the main technical, so that there is sample selection, and standardization of data processing, to determine the input and output variables to determine the hidden layer nodes, initialization of the right and threshold selection, training algorithm and parameters the selection of the final tax practice in line with the actual needs of the establishment of a network model.Finally, the thesis of Jilin Province two tax information than the background on the system, through the data warehouse technology to solve the tax analysis of staff needs, while tax revenue forecast model provides a flexible source of data, applications MATLAB neural network toolbox is established based on BP network prediction model. Predicted and actual values of the comparison to prove that this program allowed the error in the tax department can be found within the scope of tax revenue and impact the relationship between factors, can be applied in practical work to predict the future tax revenues, strengthening decision-making more scientific and normative, to improve the regulatory capacity of tax authorities and law enforcement levels, give full play to the overall effectiveness of information technology.