Mining Daily Activity Chains from Large-Scale Mobile Phone Location Data.

Area of Science:

• Urban planning
• Transportation science
• Public health

Background:

• Understanding daily human activity chains is crucial for urban planning, transportation, and public health.
• Mobile phone location data offers large sample sizes for activity pattern analysis.
• Existing machine learning models for activity inference often function as black boxes.

Purpose of the Study:

• To develop a flexible, white-box method for mining human activity chains from large-scale mobile phone location data.
• To integrate spatial and temporal features with varying weights for accurate activity pattern analysis.
• To investigate the impact of data scale on the model's performance.

Main Methods:

• A novel white-box methodology integrating spatial and temporal features.
• Weighting of spatial and temporal features to capture diverse activity patterns.
• Validation against travel survey data and a state-of-the-art method.

Main Results:

• The proposed method successfully mines major human activity chain patterns.
• Frequency distributions of identified patterns align well with existing travel survey data.
• A dataset representing 16.5% of a city's population yields reasonable outcomes.

Conclusions:

• The study provides an effective and transparent approach for analyzing daily activity chains using mobile phone data.
• The findings demonstrate the practicality of using big trajectory data for domain experts.
• The research highlights the importance of data scale in achieving reliable activity pattern insights.