摘要
获得精细尺度城市房价制图对城市发展研究及相关政策制定至关重要。然而由于过去的房价预测方法没有对多源数据进行融合,单一数据源有偏性使得房价制图及预测无法达到精细尺度。本文拟从多源空间数据融合的角度出发,通过深度学习的方法建立耦合卷积神经网络和随机森林拟合模型的武汉市城市房价预测模型,在精细尺度上模拟武汉市房价分布。实验表明预测模型可以有效地对武汉市房价做出预测,同时表明融合高分辨率遥感影像和社交媒体数据的模型能够得到比传统使用单一数据源的网络得到精度更高的预测结果。
Access to fine-scale urban house price mapping is critical to policy development and urban development research.From the perspective of multi-source spatial data fusion,the city’s urban house price forecasting model based on coupled convolutional neural network and random forest fitting model is established by deep learning method,and the price distribution of Wuhan city is simulated on a fine scale.Experiments show that the prediction model can effectively predict the housing price in Wuhan,and it shows that the model combined with high-resolution remote sensing imagery and social media data can obtain more accurate prediction results than the traditional network using a single data source.
作者
王浩天
袁强强
WANG Haotian;YUAN Qiangqiang(School of Geodesy and Geomatics,Wuhan University,Wuhan Hubei 430079,China)
出处
《北京测绘》
2019年第12期1549-1553,共5页
Beijing Surveying and Mapping
关键词
多源空间数据融合
深度学习
房价预测
空间分布
城市精细模拟
multi-source spatial data fusion
deep learning
house price forecasting
spatial distribution
urban fine simulation
