Human Travel Patterns Albert-László Barabási Center for Complex Networks Research, Northeastern...

Human Travel Patterns

Albert-László BarabásiAlbert-László Barabási

Center for Complex Networks Research, Northeastern U.Center for Complex Networks Research, Northeastern U.

Department of Medicine, Harvard U.Department of Medicine, Harvard U.

BarabasiLab.com

Tropical forest in Yucatan, Mexico

“Lévy flight” by a spider monkey

Animal MotionAnimal Motion

Vishwanatan et. al. Nature (1996); Nature (1999).

Wandering AlbatrossWandering Albatross

Wandering AlbatrossWandering Albatross

Sims et al. Nature (2008).

1 Jun

22 Jun

24 May

7 Aug

11 Jul

28 Jul

5 Jan

10 Oct

24 Nov

Basking shark

Porbeagle shark Sims et al. Nature (2008).

Random Walks Lévy flights

Human MotionHuman Motion

Brockmann et. al. Nature (2006)

A real human trajectory

Mobile Phone Users

CCNRCCNR

Data Collection Limitations

We know only the tower the user communicates with, not the real location.

We know the location (tower) only when the user makes a call.

Interevent times are bursty (non-Poisson process—Power law interevent time distribution).

ALB, Nature 2005.

CCNRCCNR

Interevent TimesInterevent Times

J. Candia et al.

A-L. B, Nature 2005.

0 km 300 km100 km 200 km

0 km

100

km

200

km

Mobile Phone Users

Two possible explanations

1.Each users follows a Lévy flight

2.The difference between individuals follows a power law

β=1.75±0.15

Understanding human trajectoriesUnderstanding human trajectories

Radius of Radius of Gyration:Gyration:

Center of Mass:Center of Mass:

Characterizing human trajectoriesCharacterizing human trajectories

Radius of Radius of Gyration:Gyration:

CCNRCCNR

Characterizing human trajectoriesCharacterizing human trajectories

Radius of Radius of Gyration:Gyration:

βr=1.65±0.15

Scaling in human trajectoriesScaling in human trajectories

0 km 300 km100 km 200 km

0 km

100

km

200

km

Mobile Phone Users

β=1.75±0.15βr=1.65±0.15

Scaling in human trajectoriesScaling in human trajectories

α=1.2

Relationship between exponentsRelationship between exponents

Jump size distribution P(Δr)~(Δr)-β represents a convolution between

*population heterogeneity P(rg)~rg-βr

*Levy flight with exponent α truncated by rg

CCNRCCNR

Return time distributionsReturn time distributions

Mobile Phone Users

CCNRCCNR

The shape of human trajectoriesThe shape of human trajectories

CCNRCCNR

The shape of human trajectoriesThe shape of human trajectories

Mobile Phone Viruses

Hypponen M. Scientific American Nov. 70-77 (2006).

Spreading mechanism for cell phone viruses

Short range infection process(similar to biological viruses, like influenza or SARS)

Long range infection process(similar to computer viruses)

Palla, A.-L. B & Vicsek (2007).Onella et al, PNAS (2007)

Social Network (MMS virus)

González, Hidalgo and A-L.B.,

Nature 453, 779 (2008)

Human Motion(Bluetooth virus)

MMS and Bluetooth Viruses

Spatial Spreading Patterns of Bluetooth and MMS virus

Spatial Spreading patterns of Bluetooth and MMS viruses

Bluetooth Virus MMS Virus

Driven by Human Mobility:Slow, but can reach all users with time.

Driven by the Social Network:Fast, but can reach only a finite fraction of users (the giant component).

What is the origin of this saturation?

Spreading mechanism for cell phone viruses

If the market share of an Operating Systemis under mc~10%, MMS

viruses cannot spread.

0 km 300 km100 km 200 km

0 km

100

km

200

km

Mobile Phone Users

Pu Wang Cesar Hidalgo

CollaboratorsCollaborators

Marta Gonzalez

www.BarabasiLab.com