特征房价空间分析及连续型深度置信网络预测

以上海为研究区域,利用数据爬虫手段搜集、整理上海市二手房交易数据,通过空间自相关分析二手房交易价格的空间效应,并使用连续型深度置信网络对二手房交易价格进行分析预测.研究结果表明:上海市二手房交易价格在空间上具有显著的自相关效应,在上海市核心区域存在高-高集聚效应,在周边区域呈现低-低集聚效应,而在核心与周边交界地区存在高-低集聚和低-高集聚的负向空间效应;特征变量对价格偏高区域的二手房交易价格解释力度较小;除中心区域外,基于连续型深度置信网络的特征变量对上海市二手房交易价格预测能力良好.

Spatial Analysis of Characteristics Housing Price and Prediction With Continuous Deep Belief Neural Network

Taking Shanghai City as the research area, data crawlers are used to collect and organize the second-hand housing transaction data, the spatial effect of second-hand house prices are analyzed through spatial autocorrelation, and the continuous deep belief neural networks are used to analyze and predict the second-hand housing prices. Research results show that the transaction price of second-hand housing in Shanghai City has a significant spatial autocorrelation effect. There is a high-high agglomeration effect in the core area of Shanghai City, and a low-low agglomeration effect in the surrounding areas. There are negative spatial effects of high-low aggregation and low-high aggregation at the junction of the core and surrounding areas. Characteristic variables have less power to explain the transaction prices of second-hand housing in areas with high prices. Except for the core areas, the characteristic variables based on the continuous deep belief neural network have a good ability to predict the transaction price of second-hand housing in Shanghai City.