Research On The Influencing Factors Of Residential Price In China Based On Elastic-Net

In recent years,house prices have gradually become an representative factor of the degree of economic and social development in a region.On the one hand,house prices will drive the growth of many other industries;On the other hand,house prices which are so high and so fast growth caused a lot of disadvantages and problems,and even affecting normal economic development and social stability.It can be said that house prices are related to people's livelihood.Studying the influencing factors of house prices have important practical significance.The residential prices,which best reflect the house prices level,are taken as the research object.Ten representative variables which related to residential prices are selected from three domain categories:economic,population and real estate.And the annual data are collected of 31 provincial administrative regions in China,and the time span is from 2005 to 2015.Based on the data,the variables with greater influence are extracted through variable selection and parameter estimation,and models are fitted for studying and analyzing.First of all,31 regions of our country are devided into 4 groups by k-means clustering:poverty regions,normal regions,medium developed regions and developed regions.The empirical results show that the 4 groups of regions are basically in line with the practical significance.To some extent,the errors caused by regional differences have been reduced.Then the 4 groups of regions are fitted the models with Elastic-Net in turn,and compared with the fitting result of Lasso and Ridge regression.The experimental results show that Elastic-Net is a compromise and improvement for Lasso and Ridge regression,and it can fit a model of more reasonable and more explanatory.Finally,this paper explains and analyzes the 4 groups of final Elastic-Net models.The empirical results show that different regions with different levels of economic development have different combinations of influencing factors,and each has different influence levels and manifestations.The specific results are as follows:(1)In poverty regions(such as Tibet,Guizhou,etc.),the variables which are positively related-to residential prices are per capita GDP,urban per capita disposable income,land acquisition area and urbanization level.The negative correlations variables include population density,The total population at the end of the year and natural growth rate;(2)In normal regions(such as Hunan,Hebei,etc.),the variables which are all positively related to residential prices in the selected are per capita GDP,real estate investment amount,population density,urban per capita disposable income and urbanization level.There are no negative correlation variables;(3)In medium developed regions(such as Guangdong,Zhejiang,etc.),the variables which are positively related to residential prices are real estate investment amount,population density,urban per capita disposable income,urbanization level and natural growth rate.The negative correlations variables include regional GDP,per capita GDP and urban consumption levels;(4)In developed regions(such as Beijing,Shanghai,etc.),the variables which are positively related to residential prices are the total population at the end of the year and real estate investment amount.The negative correlations variables are urbanization level.