摘要
随着一些城市二手住宅成交参考价机制的落地,社会各界对二手房价格的评估方法提出更高要求。为提高评估精度,反映正确评估概率,本文提出一种二手房价格区间估计方法。首先,建立城市二手房屋微观特征体系,利用神经网络分位数回归模型得到住宅价格的条件分布,再通过核密度估计得到房价的概率密度函数,进而确定置信度下的二手房价格估计区间。利用Python语言,获取成都市49129条住宅交易的真实数据,对该方法进行实证检验。模型结果表明:该方法估计结果可靠性高,区间宽度合理;与线性分位数回归模型相比,该方法估计结果精度更高,稳定性更好。基于此,本文提出加强房地产数据共享体系建设、加强对二手房交易市场的监管等建议。
With the implementation of the reference price mechanism of second-hand residential transactions in some cities,all sectors of society put forward higher requirements for the evaluation method of second-hand housing prices.In order to improve the evaluation accuracy and reflect the correct evaluation probability,this paper proposes a second-hand house price interval estimation method.Firstly,establishes a micro characteristic system of urban second-hand housing and uses the neural network quantile regression model to get the housing price.Secondly the probability density function of housing price is obtained by kernel density estimation,and then the second-hand house price under confidence is determined.Obtaining the real data of 49129 residential transactions in Chengdu with Python to test the method empirically.The model results show that the method has high reliability and reasonable interval width.Compared with linear quantile regression model,this method has higher accuracy and better stability.Based on this,this paper proposes to strengthen the construction of real estate data sharing system,strengthen the supervision of the second-hand housing market and other suggestions.
出处
《价格理论与实践》
北大核心
2021年第5期85-88,194,共5页
Price:Theory & Practice
关键词
二手房
房价
区间评估
神经网络
分位数回归
Second-hand house
housesing price
range setimation
neural network
quantile regression