Research On Visual Analytics Of Human Mobility Patterns Based On Trajectory Data

Trajectory is the spatial displacement of an object over a certain period of time,reflecting the spatiotemporal characteristics of the object’s movement.Analyzing trajectory data can help interpret the underlying movement patterns of individuals and provide insights for urban traffic prediction,transportation planning,and related fields.However,trajectory data is not only massive but also exhibits complex spatiotemporal correlations,which pose challenges for analysts in extracting and analyzing relevant movement patterns.Visualization and visual analysis,through visual encoding and graphics,can intuitively convey information that is often difficult for users to understand and allow users to explore deeper patterns and trends in the data through interactive means.Therefore,visualizing human mobility patterns has been a hot topic in transportation,communication,and other fields.This paper focuses on the visual analysis of human mobility patterns.It utilizes spatial partitioning to obtain urban area units and then employs trajectory clustering to extract relevant movement patterns of individuals,aiming to discover similar regions within the city to assist in urban planning and development.Spatial partitioning aims to simplify the research and analysis of human mobility patterns by transforming the study objects from individual entities to urban spatial regions,enhancing the semantic information behind the movement trajectories to help users understand their semantic intents.Trajectory clustering aims to explore the underlying patterns contained in human mobility patterns by identifying clusters with similar movement patterns.The work includes the following three aspects:1.In terms of urban area visualization,because area partitioning can simplify the research objects and enhance the semantic characteristics,it serves as the foundation for movement pattern analysis.Due to the scale effect of urban areas,different spatial scales of partitioning result in different functional semantic representations of urban regions.To address the limitations of single-scale visualization that cannot simultaneously capture macro and micro features of urban areas and lack interpretability in scale selection,a multi-scale urban area partitioning algorithm is proposed.It first partitions the urban areas using Voronoi diagram and then merges regions based on spatial adjacency clustering,preserving the micro features of the original partitioning while presenting the macro features of merged regions.Evaluation metrics are calculated to provide users with reference opinions to aid in selecting the optimal partitioning scale.Lastly,to discover the dynamic functional semantics of the regions,trajectory stop point semantics combined with the LDA(Latent Dirichlet Allocation)model are used to extract relevant semantics.The effectiveness of the proposed method is validated through experiments.2.In terms of visual analysis of mobility patterns,considering the challenges in parameter selection for trajectory clustering algorithms and the inability to observe local mobility patterns during the process of obtaining mobility patterns,a hierarchical clustering algorithm for Origin-Destination(OD)trajectories is proposed.OD trajectories represent trajectories composed of starting and ending points of trips.The algorithm first employs an adaptive clustering algorithm to obtain global mobility patterns and then utilizes a threshold-based ODbased K-means algorithm to further divide the trajectories into local mobility patterns.OD trajectories focus on the movement intentions of trajectories by only considering the starting and ending points.To address the insufficient exploration of mobility pattern correlations between regions in the current visual analysis of mobility patterns,region co-occurrence is extracted to construct region mobility networks and co-occurrence networks.Then,based on a graph embedding model,the city regions are vectorized,and dimensionality reduction algorithms are used to discover similar regions with mobility patterns.Comparative experiments and case studies are conducted to verify the performance and effectiveness of the proposed algorithms.3.Based on the above methods,a visual analysis system for human mobility patterns is developed.The system supports multi-scale partitioning of regions,global and local mobility pattern analysis,and the discovery of similar regions.Through the visual analysis system,three case studies and user evaluation experiments are conducted to demonstrate the effectiveness of the designed system.