Modelling daily activity patterns from mobile phone location data via deep representation learning
Authors: Xinglei Wang, Junyuan Liu, Guangsheng Dong, Zichao Zeng, Stephen Law, James Haworth, Tao Cheng
Organizations: SpaceTimeLab, Department of Civil, Environmental and Geomatic Engineering, University College London, London, United Kingdom · State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China · 3DIMPact, Department of Civil, Environmental and Geomatic Engineering, University College London, London, United Kingdom · Department of Geography, University College London, London, United Kingdom · Department of Geography and Resource Management, The Chinese University of Hong Kong, Hong Kong SAR, China · The Alan Turing Institute, London, United Kingdom
Passively collected mobile phone location data provide large-scale, longitudinal observations of human mobility but do not directly reveal activity purposes. The functional characteristics of visited locations offer useful contextual information, yet their relationship with activity purpose remains uncertain, particularly in mixed-use urban environments. We conceptualise activity pattern mining as an integrated process of representation, clustering, and interpretation, and propose the Activity Chain Encoder (ACE) for the representation stage. ACE is a self-supervised model that combines pre-trained urban embeddings with visit timing and duration and uses a Transformer to model the sequential organisation of stays. It is trained using masked activity modelling and identity-guided contrastive learning without requiring deterministic activity purpose labels. Learned daily representations are aggregated into user-level profiles, clustered, and interpreted through temporal-functional patterns and Census-derived demographic context. Applied to mobile phone app location data from London and compared with three representative methods, ACE supports the identification of six differentiated weekday activity-pattern groups characterised by distinct daily rhythms, urban functional contexts, and demographic associations. These complementary forms of evidence further support the development of empirically grounded activity-pattern personas, establishing a holistic route for deriving behaviourally meaningful population patterns from unlabelled mobile phone location data. The source code for the entire analytical pipeline developed in this study is publicly available at https://github.com/xlwang233/ACE.
Figures & tables
Figure 1: The proposed methodological framework for mining activity patterns from mobile phone data.
Figure 2: The study area.
Total number of users
112,112
Total number of stays
3,495,967
Total number of daily activity chains
785,476
Mean/median number of daily activity chains per user
7.0/7
Mean/median length of daily activity chains
4.5/4
Table 1: Descriptive statistics of the dataset after preprocessing.
Urban function
Probing prompt
Residential and Housing (RH)
A private home, house, flat, or residential neighbourhood where people sleep and live.
Commercial and Business (CB)
A corporate office, workplace, or business district for daytime professional employees.
Industrial and Manufacturing (IM)
An industrial estate, factory, warehouse, or manufacturing plant with heavy machinery.
Education and Schools (ES)
An academic classroom, school, or university campus during active daytime teaching hours.
Grocery and Convenience (GC)
A supermarket, grocery store, or local convenience shop for buying daily food and household items.
Medical and Healthcare (MH)
A hospital or healthcare centre providing medical treatment, ambulance, and emergency services.
Table 2: Semantic probing prompts used to interpret the urban functional semantics of places.
Figure 3: Illustration of the allocation of a stay’s semantic probability vector to 15-minute time buckets.
Figure 4: Comparison of the representation stages of LDA-Time Budget Allocation (LDA-TBA), Semantic Embedding of Activity Chains (SE-AC), Semantic Embedding of Multimodal Trips (SE-MMT), and the proposed Activity Chain Encoder (ACE). The methods differ in how activity-chain information is encoded and aggregated into individual activity-profile embeddings. AC denotes activity chain, and SP denotes significant place.
Figure 5: Cluster-level weekday temporal-semantic activity patterns. Each panel shows the mean visitation intensity across hourly intervals and urban functional dimensions for individuals assigned to the corresponding cluster. For the full definition of the semantic categories, see Table 2 .
Figure 6: Deviation of each cluster-level weekday temporal-semantic activity pattern from the population-wide baseline. Warm colours indicate above-baseline visitation intensity, while cool colours indicate below-baseline intensity.
Figure 7: Demographic index scores associated with the inferred home locations of individuals in each activity-profile cluster. An index score of 100 indicates parity with the London-wide reference distribution.
Figure 8: Cluster-level topic distributions derived using LDA-TBA. Each panel shows the average proportion of the 14 LDA topics within a cluster, with cluster size and population share reported in the panel title. The topics were selected using perplexity and are defined by their highest-weighted POI classes in Table A.2 .
Figure 9: Weekday temporal-semantic activity patterns of the clusters derived using SE-AC.
Figure 10: Weekday temporal-semantic activity patterns of the clusters derived using SE-MMT.
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Figure A.1: Geographical distribution of the highest-scoring urban functional category derived through semantic probing. Each raster cell is assigned the category with the highest probing score for visualisation only; the full semantic distribution is retained in all subsequent analyses.
Code
Category definition
Age group
1
Aged 15 years and under
2
Aged 16 to 24 years
3
Aged 25 to 34 years
4
Aged 35 to 49 years
Appendix
Table A.1: Classification of demographic variables used in the cluster demographic analysis.
Topic
Highest-weighted POI classes
1
Bus Stops (0.090); Hair and Beauty Services (0.064); Convenience Stores and Independent Supermarkets (0.054); Electrical Features (0.049); Fast Food and Takeaway Outlets (0.046); Restaurants (0.040); Cash Machines (0.029); Paypoint Locations (0.027); Cafes, Snack Bars and Tea Rooms (0.025); Letter Boxes (0.023)
2
Bus Stops (0.053); Cash Machines (0.046); Clothing (0.039); Restaurants (0.033); Fast Food and Takeaway Outlets (0.027); Accountants and Auditors (0.026); Computer Systems Services (0.023); Jewellery and Fashion Accessories (0.022); Employment Agencies (0.019); Travel Agencies (0.017)
3
Electrical Features (0.090); Bus Stops (0.059); Vehicle Repair, Testing and Servicing (0.058); DIY and Home Improvement (0.046); Cash Machines (0.029); Convenience Stores and Independent Supermarkets (0.028); Vehicle Cleaning Services (0.025); Petrol and Fuel Stations (0.024); Unspecified Works or Factories (0.022); Cafes, Snack Bars and Tea Rooms (0.020)
4
Employment Agencies (0.043); Financial Advice Services (0.042); Computer Systems Services (0.038); Solicitors, Advocates and Notaries Public (0.037); Unspecified and Other Attractions (0.034); Business Related Consultants (0.033); Pubs, Bars and Inns (0.025); Restaurants (0.025); Cafes, Snack Bars and Tea Rooms (0.025); Insurers and Support Activities (0.021)
5
Cafes, Snack Bars and Tea Rooms (0.068); Restaurants (0.058); Fast Food and Takeaway Outlets (0.043); Cash Machines (0.041); Bus Stops (0.039); Hair and Beauty Services (0.039); Pubs, Bars and Inns (0.037); Wi-Fi Hotspots (0.029); Convenience Stores and Independent Supermarkets (0.026); Property Sales (0.026)
Appendix
Table A.2: LDA topics and their highest-weighted POI classes. Values in parentheses indicate the topic word probabilities.
Human mobility appears highly diverse, yet much of a person's daily mobility can be explained by a small set of recurring behavioral templates, such as commuting, school-centered activities, caregiving, nightlife, or errand patterns. We present IBAD (Interpretable Behavioral Anomaly Detection), a framework that learns interpretable daily mobility templates and represents each individual as a distribution over mixtures of these templates. Rather than focusing on specific locations, IBAD characterizes activities that individuals perform across locations. This approach first discovers global behavioral templates using Latent Dirichlet Allocation (LDA), then employs a hierarchical self-supervised model to learn normal behavior of individuals from their soft behavioral templates. We also introduce a splicing benchmark that creates controlled behavioral mismatches between an individual's historical profile and injected mobility patterns. Experiments on real-world and synthetic datasets show that daily behavior can be effectively decomposed into a small number of interpretable templates. Crucially, we show that the learned behavioral archetypes transfer across distinct geographic and demographic contexts. Furthermore, IBAD maintains a robust competitive performance across all settings. For reproducibility purposes, the code is accessible at ~https://github.com/USC-InfoLab/IBAD.
Bita Azarijoo, John Krumm, Cyrus Shahabi
University of Southern California Los Angeles, California, USA
Location representations provide mobility models with fundamental information about the spatial position, functional characteristics, and relationships of places. However, existing embeddings are often dependent on mobility observations, unable to represent unseen locations, and weakly constrained to retain geographic distance. This limits their reuse across datasets and mobility tasks. To address these limitations, we propose LE4Mob, an inductive, distance-aware, and geography-derived location embedding framework for mobility modelling. LE4Mob extends contrastive language-location pre-training while introducing a distance-aware regularisation objective that encourages the embedding space to preserve spatial relationships. Pre-trained from geographic context, LE4Mob can encode rich spatial-semantic information and generate embeddings for unseen locations inductively. Its independence from downstream mobility task supervision also makes it transferable across different mobility tasks. We evaluate LE4Mob on individual-level next location prediction and population-level commuter flow generation. Experiments across multiple datasets and study areas show that LE4Mob outperforms strong baselines, with particular advantages in inductive settings and when downstream models rely directly on interactions between location embeddings. These findings demonstrate the potential of distance-aware, geography-derived location representations as reusable foundations for human mobility modelling.
Xinglei Wang, Stephen Law, Zichao Zeng +3
SpaceTimeLab, Department of Civil, Environmental and Geomatic Engineering, University College London, London, United Kingdom
Foundation models (FMs) trained on large-scale accelerometer data have been proposed as general-purpose feature extractors for health monitoring, but systematic evidence of their advantages is lacking. We present the first comprehensive evaluation of four open-source accelerometer FMs against supervised baselines covering 19 tasks across the domains of activity recognition including activities of daily living, clinical monitoring, and physiological inference. We find task-dependent performance results: supervised models remain competitive with FMs on human action recognition (HAR), with no consistent advantage for either, while selected FMs lead on fall and stress detection and are the most robust to sensor-placement variation. As frozen feature extractors, FMs are strongest for demographic inference, whereas sleep staging performance remains near chance level for all models. The internal FM representations show strong similarity across layers, highlighting potential for future FM improvements. Linear and frozen probing reveals that UniMTS provides the strongest representations and is the only FM that surpasses the supervised baselines without finetuning. Concept discovery analysis shows all models capture high-intensity activities clearly but struggle with sedentary, complex or ambiguous activities. We provide scenario-based deployment recommendations. Furthermore, we identify FM-derived activity profile inference-moving beyond fixed category classification-as a promising research direction.
Alexander Bräuer, Benjamin Cauchi, Nils Strodthoff
Carl von Ossietzky Universität Oldenburg, Oldenburg, Lower Saxony, 26129, Germany