| In the past,real estate tax was only levied on operating properties,but in order to regulate the real estate market and expand the revenue sources of local governments,on January 28,2011,the Chinese government launched a pilot real estate tax in Shanghai and Chongqing,and included eligible individual residences in the scope of collection,so as to improve the situation of the real estate market.Once the whole country begins to collect real estate tax,the collection of a wide range and a large number of real estate taxes is bound to have higher requirements for the appraisal technology,the traditional housing building assessment method belongs to a single assessment method,in the face of the characteristics of a large number and wide range of real estate tax tax base assessment is bound to meet the requirements of efficiency.Therefore,how to ensure the scientific and accurate nature of the real estate tax base assessment work,but also ensure that the tax base assessment of real estate tax can be carried out efficiently and at low cost,has become an important issue that the real estate appraisal industry urgently needs to study and solve.In view of the characteristics of property tax base valuation,it is an inevitable trend to use batch valuation method to assess the tax value base of real estate tax.Based on the theory of real estate tax base assessment,this paper first compares and analyzes several valuation methods commonly used in real estate tax base assessment,and explains that the traditional single valuation method cannot match the needs of real estate tax base assessment,so the batch assessment method is required.Secondly,by comparing the advantages and disadvantages of several common batch assessment methods,the application advantages and limitations of BP neural network model in batch assessment of real estate tax base are pointed out.Therefore,based on the traditional BP neural network model,this paper proposes to improve and optimize the BP neural network,so that it can be better applied to the evaluation of ordinary residential property tax base in the original batch evaluation advantage.In the specific process of model establishment,firstly,particle swarm optimization(PSO)is used to initially optimize the weights and thresholds in BP neural networks,and the obtained weights and thresholds are given to BP neural networks for subsequent network training,so as to establish an evaluation model based on PSO-BP neural networks.After that,the gray correlation analysis method is used to select the factors that play a decisive role in the price of ordinary residential real estate,that is,the input variables of the BP neural network are simplified.Finally,this paper selects the Nanping Group of Nan’an District of Chongqing as the evaluation object for empirical analysis,and constructs the traditional BP neural network model,linear model and semi-logarithmic feature price model on this basis,and compares it with the evaluation results of the PSO-BP neural network model,and finds that the BP neural network after particle swarm optimization has great improvement in dealing with nonlinear problems,and its convergence and generalization are also greatly improved,and its evaluation results are closer to reality and have high accuracy.And the indicators of the evaluation results are better than the current mainstream batch evaluation method-feature price model price model.The results show that the batch assessment model of ordinary residential real estate tax base based on PSO-BP neural network can estimate the tax value of ordinary residential real estate faster and more accurately,and has broad application prospects. |
