基于神经网络房地产价格指数的预测研究(续)

研究表明,房地产价格指数常表现为非线性,要对它进行预测就必须利用一种能模拟非线性的模型。从理论上讲,神经网络能够无限逼近非线性函数,所以本文便尝试采用神经网络模型作为预测的模型。本文具体运用的是基于误差反向传播算法的多层前馈网络(BP神经网络)和径向基函数(RBF)神经网络。首先利用BP神经网络对采集到的中国房地产价格指数进行训练和模拟,最后进行预测,并比较预测结果和真实值,发现误差比较大,一方面是因为选取的样本数据少,另一方面是因为BP神经网络本身具有缺陷。为了克服BP神经网络预测的缺陷,本文接着运用RBF神经网络对选取的数据进行训练和模拟,用训练好的网络来进行预测,得到的预测结果与真实值相比较,误差很小,而且RBF神经网络的运行速度要比BP神经网络快很多。经过比较可以得出RBF神经网络用于经济预测可以达到很好的效果。

Research on Forecasting Real Estate Price Index Based on Neural Networks

The research indicated that the performance the real estate price index is often a non-linearity, If we want to carry on the forecast to the real estate price index we have to use a model which is able to simulate the non-linear date. Theoretically speaking, the nerve network can infinitely approach the nonlinear function, so this article takes the nerve network model as forecast model. This article specifically utilizes the BP nerve network and the RBF nerve network. First I use the BP nerve network to train and simulate the Chinese House price index which gathers, finally carries on the forecast. Comparing forecast result with real value, I discovers that the error is quite big. On the one hand is that the sample data selected is small, on the other hand is that BP nerve network itself has some flaws. In order to overcome the BP nerve network's flaw on forecast, this article then carries on the training, the simulation using the RBF nerve network to the selection data, with the network which trained carrying on the forecast, obtains the forecast result and compare real value with forecast result,I find that the error is very small, Moreover the RBF nerve network running rate must be quickly more than the BP nerve network. After the comparison I find we can achieve good effect using the RBF nerve network to forecast in economy.