Predictability Limits in Human Dynamics as a Function of Data Completeness
Abstract
The scientific analysis of the regularities observed in individual and collective human movement trajectories is of fundamental relevance to a wide range of areasÑurban planning, wireless communication design, the prevention of epidemics, and natural security issues such as detection of clandestine activity, to name but a few. The ubiquity of mobile phones and location-based social media has enabled the capture of comprehensive time-resolved individual information, offering a unique opportunity to observe and predict human activity on an unprecedented scale. Yet, major gaps remain in our understanding of human dynamics. In particular, the limitations involved with the development of theories and frameworks with any predictive power that is applicable when presented with incomplete or partial information, i.e. sparse, missing, or corrupted data. Inspired by recent breakthroughs in our research, in this proposal we aim to fill this gap through a novel combination of machine learning techniques, statistical physics, and the analysis of dynamical social networks. Our proposal is grounded in the idea of recency as a universal phenomenon in human activity. Indeed, cognitive psychology has demonstrated a measurable enhancement in the accuracy of human beings recollecting newer events as compared to older ones. In previous work, we demonstrated that this effect strongly manifests itself in human movement due to the existence of strong biases towards the return to recently-visited locations, independent of the frequency of visit to the said location. It is also likely that there is a strong interconnection between human movement and social ties, whereby people within the same social network are known to visit similar or identical locations. While there is limited quantitative understanding of the connection between these two facets of human activity, progress can be made by measuring the effect of recency in social tie formation. That is, the probability of forming social ties in the future given recent encounters between individuals. A comparison of the recency effect in movement and tie formation may serve as a clue towards connecting the dynamics of the two activities. Consequently, the first part of the proposal will deal with measuring the recency effect in social tie formation (if any) and compare it to our discovered statistical properties of recency in mobility. The influence of recency on human mobility and the potential effect it may have in social tie formation leads naturally to the question of observation windowsÑthe temporal range that may be considered ÒrecentÓ and the optimal period of observation required to accurately capture the dynamics of the two human activities. In other words, what level of predictability can one achieve as a function of the size of the observation window of past activities? Thus, the second part of this project will determine: whether the limits of predictability of these phenomena can be improved by combining the two. That is we will determine whether the knowledge of an individualÕs mobility can be enhanced by awareness of the associated social network.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 16, 2018
- Source ID
- W911NF1710127
Entities
People
- Gourab Ghoshal
Organizations
- Army Contracting Command
- United States Army
- University of Rochester