基于线性残差变系数模型的哈尔滨房价研究

Research on housing price in Harbin based on linear residual variable cefficient model

被广泛使用的线性模型具有良好的解释性和外延性,但自适应性弱,有时拟合和预测效果欠佳;变系数模型则反之.为解决这一类矛盾,提出分段解决方法:先建立显著的线性模型,再基于线性模型残差建立其余相关影响因素的变系数模型,该方法在保持模型良好解释性和外延性的同时,提高了模型拟合和预测精度.基于哈尔滨市(包含九区三县两市)2000~2017年的住宅商品房相关数据,对哈尔滨市平均房价及其影响因素进行分析,建立基于线性残差的变系数模型,数据结果表明,这种方法在研究这类问题时优于单一的线性模型和变系数模型.

The widely used linear model has good explanation and extension,but it’s adaptability is weak,and sometimes the fitting and prediction effects are poor;the variable coefficient model is the opposite.In order to solve this kind of contradiction,a segmented solution method was proposed:firstly established a significant linear model,built the variable coefficient model of the remaining relevant influencing factors based on the residual of the linear model.This method improved the fitting and prediction accuracy of the model while maintained the good interpretability and extensiveness of the model.Based on the relevant data of residential commercial housing in Harbin(including nine districts,three counties and two cities)from 2000 to 2017,the Harbin housing price and its influencing factors were analyzed,and a variable coefficient model based on linear residuals was established.The data results showed that this method was superior to a single linear model and a variable coefficient model in studying such problems.