Modeling Individual Daily Social Activities from Travel Survey Data

摘要

Inferring activity types from the massive human-tracking data is of great importance for the understanding of human daily activity patterns in the cities. Researchers have investigated various methods to infer activity types automatically, however, the recognition accuracy of social activity types (such as shopping, schooling, transportation, recreation, and entertainment) are not satisfactory. This research proposes a machine-learning-based method to model individual daily social activities from travel survey data. Using Guangzhou as an example, we extract 21 dimensional spatial and temporal attributes to construct the random forest (RF) method to identify and validate social activities at the individual level. The experiment result shows the recognition accuracy of our approach is 75%. The effects of different factors on social activity participation are also investigated. The proposed approach can help us better understand human behaviors and daily activities, and also provide valuable insights for land use and traffic management planning and other applications.