Research On Real Estate Evaluation Based On The BP Neural Network

Our urbanization have seen the rapid development of the real estate market.Consumption,investment and other transactions are emerging in large numbers in real estate market.Among them,transactions associating with residential real estate which occupy the largest propotion of devolopment and investment are the most dynamic.Also,the rapidly increasing housing prise has aroused various of conflicts thus attract a worldwild attention from every aspect of society.It is critical to objectively and accurately assess an real estate,so that people can grasp the real estate prices and their trends.However,traditional real estate appraisal approachs consume lots of manpower and material resources and it is slow and subjective,therefore it cannot meet the large number of appraisal demand in real estate market.In addition,due to the special properties inherent in the real estate itself,the characteristic factors affecting the price of real estate are very complicated,including not only qualitative data factors but also quantitative data factors.To effectively use the informations associating with these transaction real estate,we must collect,collate and effectively store them.Therefore,it is of great theoretical and practical significance to improve the traditional method of to find a more scientific evaluation approach to study the valuation of real estate.BP neural network is a powerful,widely used machine learning algorithm,its adaptability,non-linearity and large-scale parallel processing ability enable it to significantly reduce the consumption of manpower and material resources,and to deal with nonlinear problems with high efficiency,also to reduce the subjective random,thereby to play an effective role in the real estate assessment.The real estate evaluation based on the BP neural network is mainly depend on the theory of traditional methods,but collect and process data information,find the objective law between the real estate evaluation and its influencing factors through computers instead of human.Evaluation using this method is better at Efficiency and objectivity.Besides,more and more families choose to buy second-hand housing because the exorbitant commodity housing prices beyond their ability to pay.Hence a large number of sample data of second-hand housing transactions could be obtained for the valuation of real estate.Firstly,this paper presents relevant theoretical researchs on real estate evaluation and applications of artificial neural network,and puts forward the research ideas and frame of this paper.Secondly,the paper introduces the types and characteristics of residential real estate,and systematically combs the widely used methods in the current real estate price evaluation inclding their theoretical basis,application condition,range,advantages and disadvantages.Then based on the shortcomings of traditional methods,this paper introduces the basic idea of neural network into real estate appraisal,and ana lyzes the feasibility and superiority of BP neural network in real estate appraisal.Afterwards,according to the theoretical methods in the first two chapters,this paper comprehensively demonstrates the influencing factors of real estate price,and const ructs the influencing factors index system of residential real estate price,also quantifies,normalizes and trenches the indexes.Further this paper using random forest theory to sort the indicators,then reduce the number of indicators to optimize the index system.Thirdly,combined with BP neural network theory,the real estate evaluation model based on BP neural network is constructed,and the input layer,output layer,hidden layer and network structure parameters of neural network are designed and det ermined.In addition,the model is optimized by MATLAB software.Finally,this paper collects the sample data from the real estate of the school district in Beijing,and put them into the evaluation model to train.It is found that the model can predict th e real estate value accurately,thus proves the accuracy and validity of this improved evaluation method.At the same time,in order to carry out the real estate value evaluation better,on the basis of the preliminary training model,the optimal evaluation model is determined by adjusting the parameters and selecting the function.And the superiority of the model is illustrated by comparing with the random forest method.