F-Deepwalk: A Community Detection Model for Transport Networks

Area of Science:

• Urban planning and transportation network analysis.
• Computational social science and network science.

Background:

• Transportation network design typically relies on dividing metropolitan areas into communities.
• Accurate community division is crucial for optimizing transportation routes, station placement, and inter-community travel links based on scale, population density, and travel behavior.

Purpose of the Study:

• To develop an improved method for urban spatial community detection in transportation networks.
• To enhance the classic Deepwalk model for more precise community identification and transportation network planning.

Main Methods:

• Modified the classic Deepwalk model by proposing a Random Walk (RW) algorithm with generalized travel cost and an improved logit model.
• Employed the K-means algorithm for urban spatial community detection, resulting in the F-Deepwalk model.
• Validated the model using a basic road network and the Shijiazhuang urban rail transit network.

Main Results:

• The F-Deepwalk model, incorporating generalized travel cost, demonstrated a higher profile coefficient compared to the basic Deepwalk model.
• Model performance improved as the random walk length decreased.
• The model's accuracy was further confirmed using a real-world urban rail transit network.

Conclusions:

• The F-Deepwalk model offers a more accurate approach to urban community detection for transportation network design.
• Considering generalized travel costs significantly enhances the effectiveness of community detection algorithms.
• This method provides a valuable tool for optimizing urban transportation systems to better meet resident travel demands.