Mixing Individual and Collective Behaviours to Predict Out-of-Routine Mobility (2404.02740v2)

Abstract: Predicting human displacements is crucial for addressing various societal challenges, including urban design, traffic congestion, epidemic management, and migration dynamics. While predictive models like deep learning and Markov models offer insights into individual mobility, they often struggle with out-of-routine behaviours. Our study introduces an approach that dynamically integrates individual and collective mobility behaviours, leveraging collective intelligence to enhance prediction accuracy. Evaluating the model on millions of privacy-preserving trajectories across three US cities, we demonstrate its superior performance in predicting out-of-routine mobility, surpassing even advanced deep learning methods. Spatial analysis highlights the model's effectiveness near urban areas with a high density of points of interest, where collective behaviours strongly influence mobility. During disruptive events like the COVID-19 pandemic, our model retains predictive capabilities, unlike individual-based models. By bridging the gap between individual and collective behaviours, our approach offers transparent and accurate predictions, crucial for addressing contemporary mobility challenges.

• The paper introduces a hybrid approach that integrates individual and collective mobility behaviors to predict out-of-routine movements with superior accuracy.
• It demonstrates robust performance with high ACC@5 scores in low-overlap urban scenarios, outperforming both individual-based and deep learning models.
• The study underlines practical applications for urban planning and pandemic response by effectively forecasting non-routine mobility in densely populated areas.

Mixing Individual and Collective Behaviours to Predict Out-of-Routine Mobility

Introduction to the Study

The paper introduces a novel approach for predicting human mobility by dynamically integrating individual and collective mobility patterns. This methodology addresses the challenges of predicting out-of-routine movements, which are typically difficult for models focused solely on individual behaviour. Through evaluating the model with millions of anonymized, privacy-preserving trajectories across three US cities, it demonstrates superior performance in predicting out-of-routine mobility compared to both individual-based models and advanced deep learning methods.

Evaluation and Results

The model's effectiveness is quantified using top-5 accuracy (ACC@5) across different scenarios, including low to high trajectory overlap between training and test sets. The findings reveal that:

• The proposed model outperforms individual-based models and equals or slightly surpasses the accuracy of deep learning solutions in routine prediction scenarios.
• It significantly excels in low-overlap scenarios, indicating strong predictive power for out-of-routine movements. This underscores the model's capability to integrate collective behaviour effectively, with notable improvements observed especially in areas of high points of interest density.
• Spatial analysis further highlights the model's increased accuracy in urban centres, suggesting that collective mobility behaviours have a pronounced influence in densely populated or commercially active areas.

Implications and Future Directions

This work carries considerable implications for the development and application of predictive models in urban planning, public health, and mobility management. The ability to accurately forecast out-of-routine movements can significantly enhance efforts in traffic management, urban design, and epidemic response, among other fields. The paper suggests possible future research directions, including the integration of points of interest density into predictive models and exploring the potential of hybrid models that combine the interpretability of Markov models with the predictive power of deep learning.

Reliability Under COVID-19 Restrictions

A critical aspect of the paper involves assessing the model's reliability under significant behavioural shifts, such as those induced by the COVID-19 pandemic. The findings indicate that models based on collective information, including the proposed method, exhibit stronger resilience to changes in human mobility patterns caused by the pandemic. This resilience emphasizes the utility of incorporating collective behaviours into predictive models, especially in scenarios characterized by rapid and widespread changes in individual mobility behaviours.

Conclusion

The approach outlined in this paper represents a pivotal advancement in human mobility prediction, bridging individual and collective behaviour patterns to improve accuracy and reliability, particularly in predicting out-of-routine movements. The paper underscores the necessity of considering collective behaviours in predictive models and highlights the potential of such approaches in addressing contemporary challenges in mobility and urban planning. Future work might expand on these findings by further exploring the interplay between spatial features, points of interest, and mobility patterns to refine predictions and enhance model applicability across varied contexts.