Abstract: The transaction data of ordinary residential house prices in Hefei City from 2016 to 2017 was considered. By using the spatial interpolation method and trend analysis method, the spatial changes in residential prices were analyzed. It was found that the house prices in Hefei gradually decreased from south to north and decreased from the center to the edge districts in the east-west direction. Expanding the two-phase change-point estimation method of Jin et al. and using the new change-point detection algorithm we found a change point which divided the residential price into two intervals, and we analyzed separately to establish a spatial lag model. The research results show that the residential prices in Baohe District show a strong spatial auto-correlation, and there are obvious spatial agglomeration characteristics. It is better to build a spatial lag model by finding out the change points and then separately building the spatial network auto-regressive models. There are many factors that affect house prices. Business districts, subway stations, school districts, plot ratios, and total floor area all have certain impacts on prices.

Key words: spatial auto-correlation, spatial network auto-regressive change point model, change point