Decoding Urban Mobility: Application of Natural Language Processing and Machine Learning to Activity Pattern Recognition, Prediction, and Temporal Transferability Examination
- 2 January 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (7) , 7151-7173
Abstract
Activity patterns provide valuable insights into activity-based travel demand modeling and understanding human mobility. However, challenges often arise in accurately recognizing activity patterns, predicting activity patterns, and ensuring the temporal transferability of these prediction models. This paper presents a novel approach that combines natural language processing (NLP) techniques and cutting-edge machine learning algorithms to recognize and predict activity patterns of urban residents using 10-year interval Household Travel Survey data. The proposed method based on NLP to effectively transforms activity sequence information, identifies more categories and provides clearer and more explanatory activity pattern information compared to using only K-Means $++$ for clustering. Furthermore, the study develops a correlation model between activity patterns and socio-demographic characteristics of traveler by combining various machine learning models and identifies the optimal model with positive predictive performance for both weekdays and weekends. Lastly, the temporal transferability of the model is examined by comparing activity pattern characteristics at 10-year intervals and using machine learning interpretability methods to analyze model performance, relative importance of explanatory variables, and partial dependence plots (PDPs). The results show that the proposed model possesses temporal transferability, indicating its potential for predicting refined activity demand of the population in future years. These findings enhance our understanding of urban residents’ travel and activity demand, and have potential implications for urban planning, transportation management, and policy-making.Keywords
This publication has 25 references indexed in Scilit:
- Imputation of missing time-activity data with long-term gaps: A multi-scale residual CNN-LSTM network modelComputers, Environment and Urban Systems, 2022
- Who you are determines how you travel: Clustering human activity patterns with a Markov-chain-based mixture modelTravel Behaviour and Society, 2021
- From edit distance to augmented space-time-weighted edit distance: Detecting and clustering patterns of human activities in Puget Sound regionJournal of Transport Geography, 2019
- A time-use activity-pattern recognition model for activity-based travel demand modelingTransportation, 2017
- Modeling the activity profiles of a populationTransportmetrica B: Transport Dynamics, 2016
- Model-Based ClusteringJournal of Classification, 2016
- Learning daily activity patterns with probabilistic grammarsTransportation, 2015
- Activity-based models of travel demand: promises, progress and prospectsInternational Journal of Urban Sciences, 2013
- Clustering daily patterns of human activities in the cityData Mining and Knowledge Discovery, 2012
- The usefulness of the Sequence Alignment Methods in validating rule-based activity-based forecasting modelsTransportation, 2012