| The benchmark land price as one of the main means of the government macroeconomic regulation and control of land demand,the evaluation of which is of great practical significance to standardize the town real estate market,to maintain the legitimate income of the country and to formulate urban development policy. With the rapid development of economy, the advancement of urbanization, frequent land transactions and land prices have been rising, to ensure the standard land price of real time, it has to constantly update the benchmark land price evaluation. Because of land price affected by many factors, to determine the weighing values of each influence factor and to choose benchmark land price evaluation model have some of characteristic,subjectivity, empirical and uncertainty, the traditional method of land price evaluation are often time-consuming, low intelligence and Low functional. Therefore, a different approach, to study the new evaluation model to solve the problem of the present stage of benchmark land price update work has a certain reference and reference value.The BP neural network as the core part of the artificial neural network, due to the unique learning ability and the advantage of parallel distributed information processing, which is very suitable for solving some problems of strong subjectivity, low intelligence, many factors inaccurate and fuzzy information. This determines the BP neural network is applied to the feasibility and the rationality of the benchmark land price update.In this paper, henan shangqiu 2013 benchmark land price, for example, because of the many factors influencing shangqiu benchmark land price, assume that direct use of artificial neural network to forecast the benchmark land price, on the one hand, it will bring a surge in network scale, the increase of operation time, reduce network convergence and generalization ability; on the other hand, because the correlation between the forecast factors, it lead to input information overlap, also makes the model accuracy is reduced. Therefore, first of all, with the aid of SPSS19.0 software, principal component analysis, the dimension of 20 factors get 6 combination; Second, the use of L M improved BP neural network algorithm, through Matlab2007 software building the model; Finally,organically combined with principal component analysis and BP neural network, construct the principal component analysis of BP neural network combination forecast model.In order to validate the proposed based on neural network and principal component analysis of the effectiveness of the combination forecast model, set up the following two models:(1)Don’t do any deal with the sample,directly use L- M improved BP network training and testing.(2) In front of the BP network training, using principal component analysis(pca) to extract land main influence factors, and the results as the input of the BP network to establish a network model.Through the two steps model of training time and the comparison of prediction accuracy, the results show that the combination of principal component analysis and BP neural network model training results of the average relative error was 0.46%, and the single BP neural network training results of the average relative error was 0.81%. Therefore, based on the combination of principal component analysis and BP neural network model, relatively simple structure, less input variables, short operation time, and significantly higher than the general network training efficiency and accuracy are greatly improved.At the same time, the combination of model is applied to Business services standard land price evaluation, the results are basically identical with actual land price, and the average error is only 1.66%.Through the evaluation examples, Based on principal component analysis of benchmark land price assessment of the BP neural network method is a relatively simple, practical, high accuracy of the standard land price evaluation method. |
PDF Full Text Request