Deep neural networks are increasingly utilized in mobility prediction tasks, yet their intricate internal workings pose challenges for interpretability, especially in comprehending how various aspects of mobility behavior affect predictions. This study introduces a causal intervention framework to assess the impact of mobility-related factors on neural networks designed for next location prediction -- a task focusing on predicting the immediate next location of an individual. To achieve this, we employ individual mobility models to synthesize location visit sequences and control behavior dynamics by intervening in their data generation process. We evaluate the interventional location sequences using mobility metrics and input them into well-trained networks to analyze performance variations. The results demonstrate the effectiveness in producing location sequences with distinct mobility behaviors, thereby facilitating the simulation of diverse yet realistic spatial and temporal changes. These changes result in performance fluctuations in next location prediction networks, revealing impacts of critical mobility behavior factors, including sequential patterns in location transitions, proclivity for exploring new locations, and preferences in location choices at population and individual levels. The gained insights hold value for the real-world application of mobility prediction networks, and the framework is expected to promote the use of causal inference to enhance the interpretability and robustness of neural networks in mobility applications. @en

Enhanced efforts in the transportation sector should be implemented to mitigate the adverse effects of CO₂ emissions resulting from zoning-based planning paradigms. The concept of a 15-minute city, emphasizing proximity-based planning, holds promise in reducing unnecessary travel and progressing towards carbon neutrality. However, a critical research question remains inadequately explored: to what extent is the 15-minute city concept feasible for American cities? This paper presents a comprehensive framework to evaluate the 15-minute city concept using SafeGraph Point of Interest (POI) check-in data across 12 major American cities. Our findings suggest a prevailing reliance on cars among residents due to the spatial distribution of essential activities beyond convenient walking, cycling, and public transit distances. Nevertheless, there exists significant promise for realizing the 15-minute city vision, given that most residents' daily activities can be accommodated within a 15-minute radius by low-emission modes transportation modes. When comparing cities, it appears that achieving a 15-minute walking city is more feasible for metropolises like New York City, San Francisco, Boston, and Chicago, while proving to be challenging for cities such as Atlanta, Dallas, Houston, and Phoenix. In examing inter-group comparisons, neighborhoods with higher proportion of White residents and higher median incomes tend to have more accessible POIs, with a substantial percentage of activities concentrated within a 15-minute radius. This demographic also shows a greater propensity to fulfill daily activities through walking, cycling, or public transit trips within a 15-minute travel time, thus presenting a greater potential in CO₂ reduction compared to African Americans. This study can offer policymakers insight into how far American cities are away from the 15-minute city concept. It also highlights the potential CO₂ emissions reductions that could be achieved through successful implementation.

@en

Quantifying intra-person variability in travel choices is essential for the comprehension of activity–travel behaviour. Due to a lack of empirical studies, there is limited understanding of how an individual's travel pattern evolves over months and years. We use two high-resolution user-labelled datasets consisting of billions of GPS track points from ∼3800 individuals to analyse individuals’ activity–travel behaviour over the long term. The general movement patterns of the considered population are characterised using mobility indicators. Despite the differences in the mobility patterns, we find that individuals from both datasets maintain a conserved quantity in the number of essential travel mode and activity location combinations over time, resulting from a balance between exploring new choice combinations and exploiting existing options. A typical individual maintains ∼15 mode–location combinations, of which ∼7 are travelled with a private vehicle every 5 weeks. The dynamics of this stability reveal that the exploration speed of locations is faster than the one for travel modes, and they can both be well modelled using a power-law fit that slows down over time. Our findings enrich the understanding of the long-term intra-person variability in activity–travel behaviour and open new possibilities for designing mobility simulation models.

@en

Reliable quantification of epistemic and aleatoric uncertainty is of crucial importance in applications where models are trained in one environment but applied to multiple different environments, often seen in real-world applications for example, in climate science or mobility analysis. We propose a simple approach using surjective normalizing flows to identify out-of-distribution data sets in deep neural network models that can be computed in a single forward pass. The method builds on recent developments in deep uncertainty quantification and generative modeling with normalizing flows. We apply our method to a synthetic data set that has been simulated using a mechanistic model from the mobility literature and several data sets simulated from interventional distributions induced by soft and atomic interventions on that model, and demonstrate that our method can reliably discern out-of-distribution data from in-distribution data. We compare the surjective flow model to a Dirichlet process mixture model and a bijective flow and find that the surjections are a crucial component to reliably distinguish in-distribution from out-of-distribution data. @en

Deep learning (DL) models have shown strong predictive power in solving traffic problems in the past few years. Due to their lack of interpretability and transparency, applications of such models are sometimes controversial. To ensure trust in the model, it is crucial for model end-users and decision-makers to understand how the spatiotemporal dynamics of traffic states are captured and how contextual factors are utilized by the model. Inspired by feature attribution approaches in the Explainable Artificial Intelligence field, we propose a pipeline to interpret how the UNet model, a winning DL model in the image-based short-term traffic forecast competition, learns various dependencies from data for traffic forecast. By extending the classical saliency map method, we obtain an attribution value for each element in the input data to understand its influence on the prediction. Moreover, we propose multiple aggregations over the raw pixel-level attribution results, enhancing their interpretability for traffic forecast applications. Our results show that the most recent timestamp consistently contributes the most to predicting both traffic volume and speed. The global and local spatial dependencies captured by the UNet model suggest a pattern akin to the road network, yet they differ between urban and rural contexts. Finally, we compare the contribution of static contextual factors and dynamic features. Although they share similar spatial influential boundaries, contextual factors contribute less than speed and volume features. This study provides a systematic analytical pipeline to understand how a DL model utilizes the spatial, temporal, and contextual dependencies in traffic data for traffic forecast, assisting in decision-making for intelligent transportation systems. @en

Location graphs, compact representations of human mobility without geocoordinates, can be used to personalise location-based services. While they are more privacy-preserving than raw tracking data, it was shown that they still hold a considerable risk for users to be re-identified solely by the graph topology. However, it is unclear how this risk depends on the tracking duration. Here, we consider a scenario where the attacker wants to match the new tracking data of a user to a pool of previously recorded mobility profiles, and we analyse the dependence of the re-identification performance on the tracking duration. We find that the re-identification accuracy varies between 0.41% and 20.97% and is affected by both the pool duration and the test-user tracking duration, it is greater if both have the same duration, and it is not significantly affected by socio-demographics such as age or gender, but can to some extent be explained by different mobility and graph features. Overall, the influence of tracking duration on user privacy has clear implications for data collection and storage strategies. We advise data collectors to limit the tracking duration or to reset user IDs regularly when storing long-term tracking data.

@en