Application Of BP Neural Network In Real Estate Mass Appraisal Systems

With the development of the real estate market,Mass appraisal problems produced,such as house demolition price evaluation, real estate assets approved and real estate taxation. Due to the traditional single case assessment methods exist low efficiency, high cost, easily affected by the subjective factors and other shortcomings, Which led to a loss of tax revenue, social injustice phenomenon.This paper is based on the analysis of foreign mass appraisal techniques and focus on the central part of computer mass appraisal system-the mass appraisal model.It provides theoretical support for the establishment of the system for the domestic real estate appraisal. Because of the difference between the larger countries in the real estate market, the construction of domestic mass appraisal model can not copy the foreign theories,it should be combined with the actual situation and improved by learning from the foreign mature mass appraisal technique. China’s real estate market is not yet mature, and have more complex factors.It difficult to determine the relationship between features and price variables.Therefore, the BP neural network model is proposed in this paper, by using its adaptive processing problems and nonlinear mapping ability, to find out the relationship between real estate price and its variables,for similar evaluate large quantities of real estate.This paper firstly introduces the concept and principle of mass appraisal methods, the BP neural network as the essence of mass market comparative method. Use the market comparison method has two very important key points, one is the selection and quantification of the factors; second is chosen of the case. In this paper, the influence factors of real estate price are analyzed in detail and quantitative treatment; And put forward the application of clustering analysis in evaluation of regional partition, solve the problem of choosing the comparable case. Secondly, introduce the basic theory and operation process of artificial neural network and discuss the structure and algorithm of BP neural network. According to the shortcomings of traditional algorithm, this paper is using the LM algorithm which is improvement. Through discussing the feasibility of the application of BP neural network theory in real estate appraisal, a real estate evaluation model based on BP neural network is established.Finally, an assessment of regional division as the research object, based on the influence factors of real estate prices as input,using neural network toolbox of MATLAB software, the price of real estate as an output of the BP network model, the hidden layer neurons number, results show that determined by repeated training method, is constructed in this paper the BP neural network model in after 13 iterations, to meet the requirements of precision reservation, network training. And the network convergence speed is fast, the fitting effect is good, the prediction error percentage of the test sample is superior. Therefore, the BP neural network model can be used to evaluate the mass in the region of the real estate.Finally, an assessment of regional division as the research object, a BP network model was established by using the neural network toolbox of MATLAB software which factors of real estate prices as input, real estate prices as output. The number of neurons in hidden layer was determined through repeated training method. results show that, after 13 iterations,the model is reach the predetermined accuracy requirement. And the network convergence speed is fast, the fitting effect is good, the prediction error percentage of the test sample is superior. Therefore, the BP neural network model can be used to evaluate the mass in the region of the real estate.