Predictability Limits in Human Dynamics as a Function of Data Completeness

Abstract

The objective of this project was to quantify the predictability of human dynamics from two points of view: under partial information and with the availability of additional information (e.g., using social-network information to improve the prediction of human mobility, and vice-versa). We investigated to which extent the massive availability of data, provided from a multitude of sources, such as phone call registries, social media posts,and web-browser navigation history, can aid developing algorithms such that the maximum predictability (theoretical) is reached. Here we report the following analyses: (1) the characteristics of a candidate data set, (2)the creation of social graphs using human mobility data, (3) the recency effect on social interactions, (4) the predictability in human dynamics, (5) the predictability in human dynamics under partial information, (6) the interplay between the predictability of social ties and human mobility (all previously reported), plus (7) strong evidence of circadian and ultradian rhythms to human mobility an exciting new development not discussed in the original proposal.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 14, 2018
Accession Number
AD1226293

Entities

People

  • Gourab Ghoshal

Organizations

  • University of Rochester

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Database Systems and Applications
  • Regression Analysis.