In Cities (London, England)
Understanding residents' daily activity chains provides critical support for various applications in transportation, public health and many other related fields. Recently, mobile phone location datasets have been suggested for mining activity patterns because of their utility and large sample sizes. Although recently machine learning-based models seem to perform well in activity purpose inference using mobile phone location data, most of these models work as black boxes. To address these challenges, this study proposes a flexible white box method to mine human activity chains from large-scale mobile phone location data by integrating both the spatial and temporal features of daily activities with varying weights. We find that the frequency distribution of major activity chain patterns agrees well with the patterns derived based on a travel survey of Shenzhen and a state-of-the-art method. Moreover, a dataset covering over 16.5% of the city population can yield a reasonable outcome of the major activity patterns. The contributions of this study not only lie in offering an effective approach to mining daily activity chains from mobile phone location data but also involve investigating the impact of different data conditions on the model performance, which make using big trajectory data more practical for domain experts.
Yin Ling, Lin Nan, Zhao Zhiyuan
Activity chain, Activity purpose, Data size, Mobile phone data, Trajectory analysis