Fig. 2.

Flowchart of TimeGeo and input features extracted from active CDR users. (A) Spatial and temporal choices per time step. Three individual specific parameters control temporal patterns, including the weekly home-based tours ($n_w$), dwell rate (${\beta }_1$), and the burst rate (${\beta }_2$). $n_w$ influences the travel likelihood when a person is at home, ${\beta }_1{n}_w$ influences the travel likelihood when a person is out of home, whereas ${\beta }_2{n}_w$ influences the likelihood of performing consecutive out-of-home activities. (B) $P\left(t\right)$ shown here is the empirical travel circadian rhythm in an average week measured from data for active noncommuters (who have no journey-to-work trips). (C) Joint distribution of ${\beta }_1{n}_w$, ${\beta }_2{n}_w$, and $n_w$ for active noncommuters in the CDR data set. The 2D marginal distributions are shown by the contour plots. The green dot is the most probable parameter value combination with $n_w=6.1$, ${\beta }_1{n}_w=22.4$, ${\beta }_2{n}_w=508.0$. (D) Empirical probability to visit a new location $P_{new}$ as a function of distinct visited locations S; it follows $P_{new}=0.6S^{-0.21}$. (E) Empirical probability of choosing the rank k location as a trip destination follows $P\left(k\right)\mathrm{\sim }{k}^{-0.86}$.