Fig 1.
Mobility networks of two typical users.
The circles in the map denote the recorded locations of users based on MFR. The size of each circle, as well as the associated percentage, expresses the weight of each location, which is determined by the waiting time at every location. A, User 1’s trajectory presents regular weekday commute between his workplace and his home, and a weekend route to scenic spots. B, User 2’s trajectory consists of 3 frequently visited locations: home, workplace, and entertainment, each of which is with different weights. The labels of locations (e.g., home, workplace, entertainment, scenic spots) are speculated from city Point of Interests (POIs) information, time and duration of stays, frequency of visits, and dates (holiday and what day of a week). Visualization of users’ trajectories can be queried from an interactive website http://tns.thss.tsinghua.edu.cn/humanmobility.
Fig 2.
Comparison of MFR (D1) and CDR.
A, The distribution of daily number of records N per person. B, The distribution P(ΔT) of interevent time ΔT. C, The distribution P(ΔS) of interevent distance ΔS. In B, C, red lines (MFR) are significantly steeper than blue lines (CDR), indicating that MFR provides finer spatial and temporal granularity than CDR.
Table 1.
Comparison of MRF(D1) and CDR at different sample sizes.
Fig 3.
Displacement of human mobility (D1, see S2 Fig for D2).
A, The distribution of displacement P(Δr) under time scales δ ≈ 7.5min, 30min, 2hour, 8hour, 32hour. B-F, The solid lines (green and blue) indicate a truncated power law and a log-normal distribution with best fitting parameters, respectively. The insets show the best power law fitting for the tails (from 70%ile to 96%ile). G, The variation trend of fitting parameter β with time scale δ with standard deviation as error bar.
Table 2.
Number of samples for each sampling interval.
Fig 4.
Gyration radius of human mobility (D2, see S3 Fig for D1).
A, The distribution of gyration radius rg under time scales δ ≥ 7.5min, 30min, 2hour, 8hour, 32hour. We further divided all users into 3 groups according to their final gyration radius rg(T) during the whole observation period T. B-D, show the convergence speeds of rg of different user group rg(T) = Rg ± 0.05Rg and Rg = 5km, 10km, 15km, respectively.
Fig 5.
Time related statistics of human mobility (D1, see S4 Fig for D2).
A, The distribution of waiting time Δt under time scales δmin and δ ≥ 7.5min, 30min, 2hour, 8hour, 32hour. B, The distribution of moving speed under time scales δ ≈ 7.5min, 30min, 2hour, 8hour, 32hour. C, The distribution of radius of gyration for 3 user groups with different max speeds 0 < vmax ≤ 6km/h, 15 < vmax ≤ 25km/h, 30 < vmax ≤ 60km/h during the whole observation period T. D, The distribution of moving distance for each group during morning rush hour (7am to 10am).