Research On Urban Grid Land Price Appraisal Model Based On Bp Neural Network

Urban land price appraisal is important content in land resource management. Basic standard land price is the basic form of the urban land price. It reflects the average price of the regional land, which couldn't reflect accurately the microcosmic land price in all of the urban regions. Grid land price is the average price of the grid unit. The smaller grid units, the smaller grid area covered. In this case, it could reflect the price of the entire town more accurately. In the appraisal area, It takes certain specific square grid for its basic appraisal unit, which we comprehensively use computer and GIS technology and select the appropriate assessment methods to obtain. On the one hand, grid land price enriches the land price component system, on the other hand, combined with other land price results, it plays more important role. They could provide more scientific basis for our government to manage the market and our society to understand the regional land price information. At last, the urban land will be used reasonably.Owing to the uncertainty and nonlinear properties of the relationship between grid land price and its factors, it's very hard to use certain formula to simulate their relations for grid land price appraisal. BP neural network has the characteristics of nonlinear dynamic simulation and no needing to find out the form of the model in advance. It provides an effective modeling approach for grid land appraisal.First of all, this paper analyzes the domestic and foreign research trend. It summarizes the basic theory and methods of grid land price, including its concept, factors and appraisal methods. It elaborates the learning algorithm of BP neural network, and its advantages and disadvantages. In this part, it also proposes improvement measures for BP neural network's shortcomings. In the further, this paper investigates the construction of grid land price model based on BP neural network in detail, consisting of the sample data processing, the design of the initial weights and threshold, the determination of the network structure and parameter, the model's program achievement. On the basis of the above, taking Xing'an County in Guangxi as example, it uses MATLAB to program for training the network with the help of samples. It establishes grid land price model based on improved BP neural network. In contrast, it also establishes standard BP neural network and regression analysis model, and then compares their results. The results shows that the model based on improved BP neural network has better generalization and could assess the grid land price more accurately. Finally the grid land price appraisal is summarized and looked ahead.