Abstract
Community house price forecasting has been a livelihood issue for the governments and the residents, and accurate forecast of real estimate price is so important to urban planning as well as house-purchase suggestions. However, the price of residential communities involving many aspects including economic factors, community attributes and time series trend. What's more, in this paper, we take spatial dependence among communities into account, which is hard to capture in city-level. Finally, we propose a novel deep network framework to integrate all the aspects and model the spatial-temporal patterns for community house price forecasting.