Method for the Analysis and Visualization of Similar Flow Hotspot Patterns between Different Regional Groups

Haiping Zhang, Xingxing Zhou, Xin Gu, Lei Zhou, Genlin Ji, Guoan TangISPRS Int. J. Geo Inf 2018 · DOI

TL;DR

This paper introduces a novel method for identifying and visualizing similar interaction hotspot patterns between groups of regions, moving beyond traditional point-to-point or area-to-area analyses. By defining regional adjacency and employing geo-information Tupu, the approach effectively reveals complex inter-regional flow dynamics, offering a powerful tool for understanding large-scale spatial interactions.

What Problem Does This Paper Address?

Geographers and urban planners increasingly recognize the importance of "flow space" – the dynamic interactions between places driven by people, goods, and information – in understanding urban networks and functions. While existing methods can analyze individual flows or simple area-to-area interactions, there's a significant gap in understanding how groups of regions interact, especially when these groups have predefined spatial relationships (e.g., neighboring regions). Current approaches often struggle with defining these regional group adjacencies and visualizing the complex patterns that emerge. This paper aims to fill this void by providing a method to identify and visualize similar interaction hotspot patterns between such regional groups.

Methods

The proposed method, termed RG-Flow-Pattern, involves several key steps:

  1. Data Input and Preprocessing: The process starts with node-based flow data (e.g., individual travel records), which are aggregated into region-to-region flow data using GIS tools.
  2. Regional Adjacency Modeling: A crucial innovation is the definition of regional adjacency. The paper proposes four methods:
  3. Adjacent Edges: Regions sharing a common edge.
  4. Adjacent Edges and Corners: Regions sharing an edge or a node.
  5. Customized Adjacent Range: Regions within a defined buffer zone around a target area.
  6. Logical Adjacent Relationship: Adjacency defined by custom logical rules, independent of spatial position.
  7. Region Merging: Based on the defined adjacency and a traffic threshold, adjacent areas are merged to form regional groups.
  8. Flow Pattern Mining: An algorithm identifies similar hotspot flow patterns between these regional groups. This involves defining regional group interaction hotspot flow models and calculating interaction strength (P value) and total interaction value (V value).
  9. Visualization: Geo-information Tupu theory is employed for visualization. This method allows for the representation of multiple flow patterns on a single map by classifying them into basic and complex types. Visual variables like color (warm for hot spots, cool for cold spots), saturation (contribution rate of individual regions), and arrow thickness (interaction value size) are used to convey rich information.

Datasets

The study utilizes air traffic flow data for mainland China. The data is sourced from Tencent's location big-data platform, providing daily migration data for cities. For the case study, the top 10 inflow and outflow data points from each prefecture-level city were used, focusing on flight data due to its long-distance nature. The data includes origin city, destination city, and a "hot value" representing passenger numbers, which is used as the interaction value.

Key Insights and Ideas

  • Beyond Point-to-Point: The paper highlights the limitations of traditional flow analysis methods that focus on individual origins and destinations, arguing for the necessity of analyzing interactions between groups of regions.
  • Defining Regional Adjacency is Key: The method's strength lies in its flexible definition of regional adjacency, allowing for various spatial relationships to be considered when forming regional groups.
  • Geo-information Tupu for Complex Patterns: The application of geo-information Tupu is a significant contribution, enabling the visualization and classification of multiple, complex regional group flow patterns on a single map, overcoming the limitations of traditional map representations.
  • Visualizing Interaction Strength and Contribution: The use of multiple visual variables (color, saturation, arrow thickness) provides a comprehensive understanding of not only the existence of interaction hotspots but also their strength, the contribution of individual regions, and the overall interaction volume.

Key Results

The proposed RG-Flow-Pattern method was applied to China's air traffic flow data. The results demonstrated that the method can efficiently:

  • Identify regional group interaction hotspots: The analysis successfully identified patterns of strong and weak interactions between groups of cities.
  • Visualize complex flow patterns: Figures 10 and 8 showcase examples of identified flow patterns, illustrating both "hotspot" (strong interaction) and "cold-spot" (weak interaction) patterns between regional groups. For instance, Figure 10a shows a cold-spot pattern from southwestern China to the eastern coastal areas, with color depth indicating regional contribution.
  • Distinguish pattern types: The visualization effectively differentiates between various types of regional group flow patterns, such as many-to-many or one-to-many interactions, with single or double directions.

Limitations and Future Work

The authors acknowledge several limitations and suggest future research directions:

  • Short Interaction Distances: The current RG-Flow-Pattern method is more effective for analyzing regions with long interaction distances. It faces difficulties in identifying interactions between regional groups with short distances. Future work aims to develop a model suitable for mining regional group interaction patterns with shorter distances, possibly by setting smaller partitions.
  • Self-Interactive Patterns: The method may not fully recognize "self-interactive" patterns where a single region within a starting group strongly interacts with a single region in an ending group. Further research is needed to improve the identification and visualization of these specific self-interactive patterns.
  • Customization of Area Size: While predefined regional groups are used, the authors note that interactive areas are often predefined and show geographic importance, rather than being fully customizable based on size. Future work could explore more flexible customization of area sizes.
  • Comprehensive Evaluation: While the method provides rich information for individual flow patterns, assessing the overall characteristics of all models collectively could be enhanced. The authors suggest using a four-quadrant diagram to represent the four scenarios of interaction strength and value size for better overall model assessment.

Summary updated

BibTeX
@article{zhang2018method,
  author = {Haiping Zhang and Xingxing Zhou and Xin Gu and Lei Zhou and Genlin Ji and Guoan Tang},
  title = {{Method for the Analysis and Visualization of Similar Flow Hotspot Patterns between Different Regional Groups}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2018},
  doi = {10.3390/ijgi7080328},
}

Loading related content...

Figure