Study And Application Of Artificial Neural Network Model On Real Estate Appraisal

With the cancellation of the housing benefits system and the housing distribution of the implementation of monetary, real estate market has become increasingly active. The real estate industry has become an important part in the national economy, and the real estate appraisal business came into being, and the rapid development and growth, It has played an increasingly important role in promoting the normalization of real estate prices to protect the real estate fair trade, and establishing a healthy real estate market system and a fair and rational trade order. The study and application of real estate appraisal approaches is still in the exploratory stage, the existing approaches focus experience and art, and very little to establish a mathematical model of quantitative appraisal, So finding scientific and rational approaches for quantitative appraisal has practical significance and theoretical significance.The thesis simply summarized the basic theory, the basic formula and the appraisal steps of the existing domestic three basic appraisal approaches——market comparison approach, cost approach and income approach. On the basis, deeply analysed the characteristics, limitations, and relevance of these different kinds of approaches, reveals the deficiency of the traditional appraisal approaches on reflecting the nonlinear relationship between impact factors of real estate and its price, and showed the potential, difficult to avoid subjective arbitrariness in the use of these appraisal approaches.According to the characteristics of real estate appraisal, in the basis of previous studies use neural networks to deal with highly nonlinear dynamic problems, introduces the theory to the field of real estate appraisal, and created real estate appraisal model based on neural network, Using super self-learning ability of the neural network, from a large number of transaction cases to discovered and mastered the complex objective laws between real estate prices and its impact factors.In the realization of neural network real estate appraisal model, using the functions provided by MATLAB neural network toolbox designed and optimized BP network and RBF network on the application of real estate appraisal. Using the true cases by market research trained and validated the two networks, both the network models have been good forecast results. On the basis of analysis on the training and validating results, comparatively analysed the two neural network models, and pointed out advantages and disadvantages of each network and improvement measures.Finally, on the basis of ideas on introducing neural network to the field of real estate appraisal, advanced the creation of real estate appraisal system, development professional visualization software to improve the efficiency and reduce the cost of appraisal, in order to serve real estate development, operation and management better.