Local General Budgetary Revenue Forecasting Model

Local government revenue is not only an important part of national financial revenue, but also a significant guarantee for different levels of regional governements to put their financial functions into use. It is of great importance to establish a scientific and resonable forcasting model of regional financial revenue, to improve the precision of forcasting and to provide more valuable data for governments to make budgets and financial expenditure plans. At present, the ability of different regional governments on establishing forcasting model of financial revenue varies. Due to the differences of industrial structures and economic developments of different regions, their forcasting models can not be shared by each other. While there are not many researches about forcasting models of financial revenue and still many relevant researches waiting to be made in Zhejiang province.As a result, this paper, combining the contents and structures of regional financial revenue and fully taking into consideration the economic and other influential factors of financial revenue, according to the principle of integrating quantitative and quantitative analysis, makes use of traditional time sequencial method, multi-meta regression as well as SVM to establish single forcasting models for general budget revenues and several main tax types. It combines the above three methods by introducing combining-forcasting methods effectively, and establishes a reletively integrated forcasting model of regional financial budget revenues on the basis of real data of Zhejiang province as well as forcasts the general budget revenue of regional finance from 2006 to 2010 in Zhejiang province.The major contributions of this paper are it establishes a scientific and resonable revenue forcasting model for regional finance which is valuable for different governments to refer in establishing revenue forcasting models and choosing forcasting methods. According to the compareing analysis of forecasting error of different methods, concluded the conclusion that simple exponential smoothing method is costly and effectively compared with the complicated methods;while the regressional method is steady and could supply structural gap of exponential smoothing method;when the structure of the time-series is not very steady-going, combining several different methods could decrease the forecasting error.