Hedonic住宅特征价格模型的BP神经网络方法

房地产在金融市场中占有举足轻重的地位,其价格变化对整个金融市场有着显著的影响。采用特征价格模型,对美国一线城市2007年6月及2008年的房价进行了相关定价研究。对传统特征价格模型的属性因子进行了扩充,加入房产周边犯罪率因子进行模拟;在数值方法计算方面,首先对数据进行了Box-cox变换,分别采用BP神经网络及传统的最小二乘法进行数值模拟分析,结果表明,房价随犯罪事件类型及发生距离房地产的远近有-5.78%～2.08%的变化;在2008年与2007年6月的不同时段内,犯罪率的变化对房价的影响有所不同。BP神经网络模拟的价格与实际交易价格曲线比传统最小二乘模拟的价格曲线精度高出5.74个百分点。

Hedonic Housing Price Model Via BP Neural Network

In this paper,hedonic pricing model is used to assess the housing price in Washington,USA.For the pricing model,in this paper,the crime variables around the house are included.The model is built by hedonic pricing method through using traditional OLS method and neural network to simulate and with data modified by Box-cox transformation.The result shows the change in criminal rate makes the housing price change,and as the distance of crime to the housing and the types of crimes changes,the house price changes from-5.78% to 2.08%.In July of 2007 and the whole 2008,the influences of crime on housing price are different.It also shows that neural network is more accurate than the traditional OLS method with 5.74% higher degree of approximation,and shows better features.