In recent years, the theory of flow space, which measures the relationship between cities, has become an important supplement to the traditional urban system theory based on hierarchical scale standards. In the period when the flow data is relatively difficult to obtain, constructing a relatively scientific gravity model (model form, coefficient) is the key factor to rationally describe and plan the urban linkage system. Among the various types of actual flow data, the daily migration status between cities is one of the most important data resources. This data resource can help us understand the characteristics of urban network connections and put forward some thoughts on the construction of gravity model. This paper first uses the simple program written by Python to climb the daily population flow data of China’s prefecture-level units from Tencent’s location big data, and builds the urban OD network of 338 prefecture-level units. With the help of Ucinet, a tool for complex network analysis, the daily communication network of cities and towns is hierarchically analyzed at three levels: point (cities), line(flows) and surface(groups). The results show that there are great differences in the characteristics of network connections according to the level of regional development. The degree of network connectivity in coastal developed areas are relatively high, while the development level of central cities like Chengdu and Chongqing in the southwestern region are generally higher than that in general cities. The lower degree of networking in the northwestern region has made some large and medium-sized cities have strong control over the rest of the cities. The population flow data was transformed from the urban level to the provincial level. Using the Concor algorithm, it was found that the group formed by the urban network has a close relationship with the urban location, but the location restrictions on the southern cities are weaker than those in the northern cities. In general, the network connection characteristics show that the urban network connection in China is the result of the comprehensive effects of regional differences, distance attenuation and grade differences. Secondly, using the existing data, the economic factors, population size factors and distance factors in the classical gravity model are divided into regions, grades and distances. The results show that the influence of local economic development level on population mobility is U-shaped, and the urban economic growth in developed and underdeveloped regions will lead to a larger population flow; the attenuation factor in Central China is the weakest, and the possible explanation is that located in the national railway hub. The higher the city level, the stronger the influence of economy, population and distance, indicating the polarization effect of big cities. The closer the flow distance is to the urban traffic flow, the greater the hindrance of the distance. At a flow distance of more than 1 000 km, the obstruction effect of the distance is not obvious. It shows that the attenuation of the distance may only be effective in a limited space-time distance, because people are doing daily activities. Long-distance town connections may be specialized in business activities, which are less affected by time and space. The conclusions of this paper indicate that for urban space, it is necessary to pay attention to its heterogeneity characteristics at different scales and different levels and different regions. Considering the influence degree of different factors, there will be great differences. In the planning practice, it is necessary to pay attention to the construction of the theoretical model and the actual data, or use the actual data to modify the theory to obtain relatively scientific and reasonable conclusions.