Mining Person-Centric Datasets for Insight, Prediction, and Public Health Planning

Abstract

In order to increase the accuracy and realism of agent-based simulation systems, it is necessary to take the full complexity of human behavior into account. Mobile phone records are capable of capturing this complexity, in the form of latent patterns. These patterns can be discovered via information processing, data mining, and visual analytics. Mobile phone records can be mined to improve our understanding of human societies, and those insights can be encapsulated in population models. Models of geographic mobility, travel, and migration are key components of both population models and the underlying datasets of simulation systems. For example, using such models enables both the analysis of existing traffic patterns and the creation of accurate simulations of real-time traffic flow. The case study presented here demonstrates how latent patterns and insights can be (1) extracted from mobile phone datasets, (2) turned into components of population models, and (3) utilized to improve health-related simulation software. It does so within the context of computational epidemiology, applying the Data Science process to answer nine specific research questions pertaining to factors influencing disease spread in a population. The answers can be used to inform a country’s strategy in case of an epidemic.