Exploring Human Mobility Patterns: A Novel Approach to Sequential Pattern Mining and Similarity Measurement

Abstract

Understanding human mobility is crucial for fields such as public health, transportation, and urban planning, especially with the growing availability of trajectory data from smart infrastructure and location-aware devices. This paper addresses two key research questions: how to effectively discover and represent human mobility patterns, and how to measure similarity and variation among individuals and within an individual's daily life. We introduce the Hierarchical Generalized Pattern Mining (HigPam) method, which generates hierarchical generalized patterns, incorporating sequence pattern lengths and essential gap information. Additionally, we propose the iterative longest common sequence (I-LCS) method to robustly measure the similarity of daily mobility behaviors and interpersonal similarities. Combining HigPam and I-LCS, we present a comprehensive framework for examining human mobility patterns at individual and aggregated levels, illustrated through a case study in Kampala, Uganda. We hierarchically extracted activity patterns, enabling a thorough examination with varying pattern lengths. Individuals were grouped based on interpersonal similarity, resulting in four distinct groups: In Group 1 ("stay-at-home"), individuals tend to stay close to their residential areas, focusing on home-related tasks. Group 2 ("unoccupied") consists of individuals centered around their residential areas but engaging in a variety of additional activities (e.g., shopping or healthcare visits). Group 3 ("education-oriented"), with members regularly attending schools or universities. Group 4 ("work-oriented") is characterized by a predominant focus on work-related activities. This framework offers a novel approach to discover, represent, and measure human mobility patterns, providing valuable insights for applications in public health, transportation, and urban planning.