The spatial distribution of activity points depicts the fundamental state of spatial interaction. Figure 6 Spatial distribution of activity points. 4.3.2. Spatial Interaction With reference to the Shanghai Fourth Comprehensive Traffic Investigation, the city territory y-secretase inhibitor of Shanghai was divided into 35 traffic macrozones. The identities of the 35 macrozones
and the identity of the study area together constituted the item set M in the frequent item set mining. The minimum support threshold pmin was set to be 2%. The spatial interaction of residents’ activities is fetched from the outputs. The frequent 1-item sets depict the spatial distribution of activity points in different macrozones, which yields a similar result as Figure 6. Figure 7 illustrates the outcomes of frequent 2-item sets and shows the spatial interaction between two different macrozones. Figure 7 Spatial interaction of residents’ activities in the study areas. 4.3.3. Discussion Through the visualization of calculation outcomes, a brief analysis can be carried out to discover some representative features in spatial interaction. As shown in Figure 6(a), the spatial distribution of Gucun residents’ activities is a nonuniform
distribution shaped like a binuclear dumbbell. There are two centers of activity: the regional center nearby and the area in the central city along Metro line 7. As shown in Figure 7(a), both of the two centers have strong association with the surrounding areas. There also exists a strong link between the two activity centers, which plays the role of handle that joins
the centers. Figure 6(b) shows a less centripetal tendency for the residents’ activities in Dahua. The spatial distribution of residents’ activities shapes like a ribbon along Metro line 7. However, as Figure 7(b) illustrates, there are still two activity centers. Due to the short distance between Dahua and the central city, the two activity centers are closely interlinked and fuse to form one morphologically. But from the viewpoint of function level, they are still divergent. The activities of residents Entinostat in Jing’an distribute evenly without evident centralization, characterized by the flexible shape and the uniform distribution in Figure 6(c). The spatial interaction in Figure 7(c) only shows the strong associations between Jing’an and the surrounding areas. The above analysis proves the rationality of the framework proposed in this paper. The long-term and pervasive monitoring of activities based on mobile phone data is an effective way to obtain the spatial interaction between the different areas. The representative features extracted can be applied in the further studies on the interaction between individual behavior and urban space structure. 5. Conclusion Mobile phone data can pervasively track individual behavior in both temporal and spatial dimension.