| Housing is a necessity in people's life.The discussion on housing price has been popular for many years.The problem of housing price is fundamental to people's livelihood.With the rapid development of China's economy,how to find a balance between people's living standards and the increasing housing price has become the topic that public opinion attaches great importance to.The issues on similarities and differences of housing price in different regions,which factors causing high housing price and the long-term trend of housing price have become the main content of academic discussion.Studying the influencing factors of housing price and the spacetime characteristics of housing price's change can not only macroscopically help relevant departments formulate or adjust relevant policies from a more scientific perspective,but also reveal the possible changing law of housing price and the driving forces of influencing factors from a micro view,which helps to understand the real estate market much better.According to the average price of 692 commercial residential buildings in 24 periods in Nanchang City,from July 2016 to June 2018,this study uses the geographically weighted regression method to establish the regression model of housing price,and explores the spatial differentiation of housing price's influencing factors.Meanwhile,BP neural network and random forest algorithm are used to fit the house price model,calculate the importance of each feature,and predict the spatial distribution pattern of housing price.This paper then uses the space-time cube to explore the spatial-temporal distribution characteristics of housing price in Nanchang and the changes of influencing factors.The main conclusions of this paper are as follows:(1)The model of housing price based on geographically weighted regression shows that all selected factors have certain impact on house price.Location,natural landscape,convenience of surrounding facilities and some other factors have a obvious negative impact on housing price,while greening rate,property fees and some other factors have a positive impact.Different influencing factors show some differences in spatial distribution.(2)Compared with BP neural network,the forecasting model based on random forest shows better effects,which is suitable for simulating and forecasting the new housing price data set.The prediction results of spatial distribution pattern show significant spatial differentiation characteristics,showing a high-value aggregation pattern along the Ganjiang River,while the areas far from the Ganjiang River are lowvalue.Meanwhile,the fitting results of housing price in the old urban areas have relatively small errors,and relatively large errors are distributed in some emerging developing areas.(3)According to the forecasting results of the importance of influencing factors,the level factors of buildings,such as floor area,volume ratio and total number of buildings,have little influence on housing price,and the feature importances ranked from top to bottom is below: the distance between the real estate and Ganjiang River,the region to which the district belongs,the plate rating and the accessibility to subway stations.And the importance of spatial characteristic variables accounts more than 78%,which indicates that the spatial differentiation of housing price in Nanchang is mainly explained by spatial characteristic variables.(4)The housing price in Nanchang have very distinct spatial-temporal characteristics.The hot spot pattern is dominant in the area along the Ganjiang River and in the central city,while the cold spot pattern is dominant in the outer-ring of the city.And the intensity of space-time hotspot area has never weakened,but the spacetime cold point area has gradually disappeared.which means the housing price of Nanchang will rise in further period. |
PDF Full Text Request