Investigating seniors’ daily activity patterns (DAPs) is essential to understand their activity-travel needs. Although some studies have applied machine learning (ML) to derive DAPs, few of them have sought to improve the interpretability of ML. This study aims to predict and interpret seniors’ DAPs in the Chinese context by using ML with a surrogate model. First, a boosted C5.0 algorithm was employed to model seniors’ DAPs, which provided more accurate predictions than the multinomial logit (MNL) model. Second, a rule-based C5.0 algorithm was used as a surrogate model to approximate the prediction function of the boosted C5.0 algorithm and to provide insight into the underlying decision processes in the boosted C5.0 algorithm. The results show that retired men are most likely to lack out-of-home activities. A good residential built environment, especially good walkability and public transit accessibility, increases seniors’ out-of-home activities. This study provides recommendations for increasing seniors’ mobility.
Application of machine learning with a surrogate model to explore seniors’ daily activity patterns
Transportation Letters ; 14 , 9 ; 972-982
2022-10-21
11 pages
Article (Journal)
Electronic Resource
Unknown
Learning daily activity patterns with probabilistic grammars
Online Contents | 2015
|Learning daily activity patterns with probabilistic grammars
Online Contents | 2015
|Le dossier - Juniors - seniors. <>
Online Contents | 2006
Generation of Synthetic Daily Activity-Travel Patterns
Transportation Research Record | 1997
|