Modelling Large Scale Behavioral Data: Role of External Environments

Peter Edsberg Møllgaard

Research output: Book/ReportPh.D. thesis

15 Downloads (Pure)

Abstract

Understanding human behavior is a vital tool for societal progress. It plays a central role in healthcare and social science, shedding light on the complexities of well-being, culture, and interaction. This thesis takes an engineering perspective, leveraging extensive datasets to extract actionable insights into human behavior within different environments and contexts. We embrace this topic in three parts.

The first part investigates changes in human mobility patterns during the COVID-19 pandemic using data from Danish mobile network operators and the Facebook Data-For-Good project. The study identifies distinct types of mobility related to workdays, weekends, and holidays, and how these were affected by travel restrictions.

The second study addresses the impact of nocturnal mobile app usage on next-day behavior, drawing from a global dataset of 41,000 individuals and 11 million nightly observations. We find that interrupted sleep due to mobile phone use is associated with increased phone usage and decreased physical activity the following day. This research, covering data from 158 countries, reveals that 7% of adults use their phones during sleep regularly for significant durations, highlighting the importance of understanding the delayed consequences.

The third research presents a comprehensive and computational method for transforming continuous GPS signals into global stop locations with added semantic information. This process involves mapping out locations such as homes, workplaces, and Points of Interest (POIs) using OpenStreetMap data. Our analysis resulted in the identification of 2.7 billion stops across 7.1 million users. Additionally, we investigate the predictability of these semantic enhanced trajectories utilizing a transformerbased model. This model is specifically designed to collect and analyze context from diverse user behavior patterns, with the aim of predicting the next semantic element in mobility trajectories.

By examining various facets of human behavior across different environments, this thesis aim to offer a nuanced understanding of how external factors shape our actions and responses.
Original languageEnglish
PublisherTechnical University of Denmark
Number of pages128
Publication statusPublished - 2023

Fingerprint

Dive into the research topics of 'Modelling Large Scale Behavioral Data: Role of External Environments'. Together they form a unique fingerprint.

Cite this