The Research Of Rough Neural Network Used In Mass Appraisal For The Basement Of Real Estate Tax

In recent years,the real estate industry develops fast in China.It improves the living conditions of residents,drives the rapid economic growth and accelerates the urbanization.However,at the same time,with the speculation at home and abroad,the injection of capital investment,and the growing rigid demand of residents,the price of house increases too fast.The irrational growth of the real estate industry results in potential problems of economic imbalances,social instability and financial risks.And it causes threat to the real estate industry’s own sustainability.At this time the government needs to take proper controls to avoid large-scale market failures occur.However,the rigid control policies will cause the real estate bubble burst rapidly,and lead to recession,even lead to economic crisis and undermine the social stability.So the Tax Policy will more suit for the requirements of the market economic order.The core issue of full beginning of property tax is to build a viable,efficient and equitable property tax base assessment system.Mass appraisal has been the first choice of the property tax base assessment.The scholars of our country gradually introduces foreign mass appraisal of property tax base assessment to China.At this stage,the property tax base mass appraisal mainly uses market method,cost method and income method as modeling basics of mass Appraisal.There are complex,nonlinear relationship between real estate price and its influence factors.So in this paper,with a strong nonlinear mapping ability,the BP neural network is used as a technical method of the real estate price assessment.Aiming at the disadvantages of slow convergence speed and easy to fall into local minimum,the momentum adaptive learning rate adjustment algorithm is used.For the problem of redundant network,the paper uses the rough set theory to simplify and build the neural network model of rough set.In the empirical part,the paper takes the north Wu-si plate of Fuzhou as an example.On the base of analysis of factors influencing the price of real estate,it Selects 23 variables which will influence the price,and deletes the redundant condition attributes by rough set theory.And then,it models the simplified data by BP neural network and forecasts the real estate price by the established model.The simulation results show that RS-BPANN has a very good reliability in predicting prices,and has a better performance,higher prediction accuracy than ordinary BP neural network model.The fifth chapter is conclusions,recommendations and outlook.Based on the conclusion of full text,it analyzes the inadequacies in this article and gives some suggestion.