Geographically and temporally weighted regression (GTWR) has been developed to model both spatial and temporal non-stationarity in real estate market data.GTWR integrates both spatial and temporal information in the weight matrix to capture spatial and temporal heterogeneity,while the factor effects of the neighboring housing units (or zones) are totally ignored.On the other hand,a local linear fitting method (LLFM) that accounts for spatio-temporal heterogeneity in a regression context only considers distances in the factors space,ignoring the space-time locations of the neighbors and the relative space-time distance between neighbors.In this paper,we proposed a new weight function that combines the space-time distance and the distance in the factors space.A case study in Shenzhen showed that the proposed model with the new weight function performed better than the traditional GTWR or LLFM.