1. Non-Markovian Character in Human Mobility: Online and Offline 报告人：蔡世民 合作者：赵志丹，卢扬

2. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 2

3. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 3

4. 2016-6-29 互联网科学中心 4

5. 2016-6-29 互联网科学中心 5 Animal Motion Wandering Albatross Pókmajom Sirályok repülése

6. 2016-6-29 互联网科学中心 6 Sims et al. Nature (2008). Basking shark

7. 2016-6-29 互联网科学中心 Porbeagle shark Sims et al. Nature (2008).

8. 2016-6-29 互联网科学中心 8 A real human trajectory

9. 2016-6-29 互联网科学中心 9 Human Motion Brockmann et. al. Nature (2006)

10. 2016-6-29 互联网科学中心 10 Mobile Phone Users Song et al. Science (2010).

11. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 11

12. 2016-6-29 互联网科学中心 12 Metrics The trip distance distribution The radius of gyration The number of visited location over time The motif of mobility The inter-time/step distribution The dwelling time for a location The entropy and predictability ………… Used in the MS

13. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 13

14. 2016-6-29 互联网科学中心 14 Data WebsiteTower Termination The massive data set used is from a large-scale real communication system based mobile termination, where participant browses websites when they are moving from one place to another. Thus, it includes two groups of records, online trajectory of browsing websites and offline trajectory of visiting towers. In order to improve the quality of trajectory reconstruction, we remain these participants whose trajectory achieves at least 20 distinct websites and towers, and the average visiting times to each website or tower is more than 10. Finally, the total number of participants is 8,929.

15. 2016-6-29 互联网科学中心 15 Definition of Method The visual network, describing the transiting process of nodes, of which the nodes denote websites or towers, the link and vertex weights indicate the transiting and visiting numbers, respectively The dwelling time, describing consecutively visiting same node, indicates whether there is a non-Markovian character in human mobility The entropy and predictability, describing the uniformity of system and the degree of predictability, is computed based on Lempel-Ziv data compression algorithm and Fano’s inequality

16. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 16

17. 2016-6-29 互联网科学中心 17 Visual network

18. 2016-6-29 互联网科学中心 18

19. 2016-6-29 互联网科学中心 19

20. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 20

21. 2016-6-29 互联网科学中心 21 The exploration, describing the first visit to a new node, is with ρ-0.49/0.51 and λ=0.39/0.22, determined by statistical analysis of all participants. The preferential return, describing revisit the previous nodes, where each node have the probability to be revisited. The inetia, describing the dwelling time on a certain node, is simulated by excited random walk (much more details see T. Antal and S. Redner, J. Phys. A, 2005, 38: 2555)

22. 2016-6-29 互联网科学中心 22

23. Outlines Dynamic pattern of mobility Metrics for mobility Data and method Empirical results Description of model & simulating results Conclusion 2016-6-29 互联网科学中心 23

24. The rank distribution of nodes in visual network shows a truncated power-law formula at individual level and power-law one at collective level, of which the heterogeneity suggest the non-uniformity of dynamic process of human mobility. The non-Markovian character observed from both online and offline cases is suggested by the scaling law in the distribution of dwelling time at individual and collective levels, respectively. the lower entropy and higher predictability in human mobility for both online and offline cases may origin from this non-Markovian character. Yet, the distributions of individual entropy and predictability show the different degrees of non-Markovian character from online to offline cases. A protype model with three basic ingredients, preferential return, inertial effect, and exploration is proposed to reproduce the dynamic process of online and offline human mobility. 2016-6-29 互联网科学中心 24

25. 2016-6-29 互联网科学中心 25 Thank you !