Synthetic multi-day activity-travel schedules for Swedish residents

About 

This dataset contains multi-day activity-travel schedules for over 263,000 individuals residing in Sweden, representing approximately 2.6% of country's population. The individuals and their daily schedules are derived from mobile phone application data covering seven months in 2019. Mobile phone application data, one example of emerging mobility data sources, offers an alternative to other data collection methods. This data is collected by capturing phone users' geographical locations with their consent as they interact with various mobile applications.  

This open data repository includes activity-travel schedules for each individual over five simulated average weekdays, incorporating daily variability at the individual level. Each simulation day provides:  

Anonymized Identifiers: Unique IDs that link individuals across all simulation days. 

Activity Locations: Locations for home, work/school and other activities. 

Daily Activity-Travel Schedules: Detailed information on activity sequence, type, start and end times, and locations. 

Background 

The activity-travel schedules were created using a novel generative model that synthesizes individuals' average weekday activity-travel schedules from mobile phone application data. Mobile data provides geographically and population-wise extensive observations over extended periods, offering valuable insights into individuals' whereabouts. However, these datasets often include sampling biases in the population coverage and individual-level data sparsity due to intermittent and irregular phone application activities, from which the underlying geolocation data were passively collected.  

The generative model combines mobile data with the Swedish national travel survey [1]. The model employs state-of-the-art primary activity identification methods to infer individuals’ primary activity locations, i.e., home and work/school snapped to buildings. The proposed model can generate multiple schedules for each individual, showing activity sequences, types, start/end times and locations, incorporating daily variability in specific schedule attributes. At the individual level, variations occur across all elements of activity schedules, i.e., activity sequences, type, start/end times, and other activity locations, while maintaining the residential and workplace locations. Moreover, the model calculates a weight for each individual based on their residential location and inferred employment status, addressing sampling biases and ensuring a representative sample of the Swedish population.  

The performance of the generative model is evaluated by comparing its synthesized activity-travel schedules with those from the SySMo model [2], large-scale agent-based model of Sweden and with underlying travel survey data. The results demonstrate that the proposed model effectively addresses biases and sparsity in mobile phone application data, resulting in realistic and reliable activity-travel schedules. The pre-print paper "Mobile Phone Application Data for Activity Plan Generation" details the model's methodology and evaluation.

Data Description 

The current data covers 5 data files, each showing a simulation day. 

Column 

Description 

Data type 

Unit 

PId 

Unique Anonymized Identifiers 

Integer 

- 

employment  

Employment Status (0 = Not Employed, 1 = Employed) 

Integer 

- 

weight 

Weight showing the representativeness of the individuals 

Float 

- 

act_id 

Activity index of each agent 

Integer 

- 

act_purpose 

Activity purpose (work/ home/ other) 

String 

- 

act_start 

Start time of activity in minute (0-1439) 

Integer 

minute 

act_end 

End time of activity in minute (0-1439) 

Integer 

minute 

point_x 

Coordinate X of activity location (SWEREF99TM) 

Float 

meter 

point_y 

Coordinate Y of activity location (SWEREF99TM) 

Float 

meter 

point_lat 

Latitude of activity location (WGS 84)  

Float 

degrees 

point_lng 

Longitude of activity location (WGS 84) 

Float 

degrees 

Privacy Policy 

The data underlying this study were purchased from PickWell and are subject to restrictions due to licensing and privacy considerations under the European General Data Protection Regulation (GDPR). Therefore, these data are not publicly available but can be requested for research purposes through commercial access. We adhere to the guidelines established by the Chalmers Institutional Review Board (IRB) following the Swedish Ethical Review Act (2003:460) and GDPR 2016/679. The dataset contains no personal information traceable to individuals. Geolocations in this dataset are synthesized from empirical mobile application data, ensuring privacy while retaining their utility for studying mobility behavior and simulating large-scale travel demand. 

Acknowledgement 

This research is funded by the Swedish Research Council Formas (Project Number 2018-01768). The authors acknowledge Sonia Yeh for her intellectual contributions to the study. Additionally, the authors sincerely thank Jorge Gil for providing the mobile phone application data.