Daily activity-travel pattern identification using natural language processing and semantic matching

The generation of daily activity patterns (DAPs) has gained considerable attention due to its capacity to capture the interdependencies among activities and underlying behavioural dynamics. Existing clustering methods often face limitations related to the aggregation of heterogeneous DAPs, leading to reduction in prediction accuracy. This study presents a novel hybrid approach that integrates “direct activity sequence” recognition with a sequence matching algorithm grounded in Natural Language Processing (NLP) techniques. Unlike traditional methods, our approach preserves the distinct identity of each activity sequence, assigning DAPs based on the frequency of distinct sequences within each representative DAP group. This process is further enhanced by a hierarchical activity categorization structure, enabling deeper exploration of in-home activities, household interaction effects, and spatial changes between activity types. Additionally, the introduction of weighted activity categories and a match score calculation system opens new possibilities for future sequence alignment methodologies. Using data from 1808 households (approximately 6500 individuals) in Bidhannagar, India, we demonstrate that our approach outperforms traditional methods in terms of prediction accuracy. This study also explores the impact of evolving patterns of online activities, socio-economic heterogeneity, built-up area and neighbourhood characteristics, distinction between weekday and weekend on DAP prediction in the context of emerging countries.