Hierarchical Agglomerative Clustering of Bicycle Sharing Stations Based on Ultra-Light Edge Computing

Area of Science:

• Urban mobility
• Shared economy systems
• Edge computing

Background:

• Bicycle sharing systems (BSSs) are evolving into urban sensor networks.
• Operational costs, primarily from rebalancing, challenge BSS viability.
• Rebalancing currently uses clustering based on station behavior and proximity.

Purpose of the Study:

• To propose a novel clustering algorithm, HAC-ULECA, for BSS rebalancing.
• To focus clustering on station behavior, excluding proximity.
• To enable edge computing for real-time analysis within the Internet of Shared Bikes (IoSB).

Main Methods:

• Developed Hierarchical Agglomerative Clustering (HAC) integrated with an Ultra-Light Edge Computing Algorithm (ULECA).
• ULECA models station behavior using net arrivals/departures, reducing computational needs.
• Applied HAC-ULECA to real-world data from Madrid's BiciMAD system.

Main Results:

• HAC-ULECA effectively separates docking stations based on distinct behavioral profiles.
• The algorithm identifies spatio-temporal patterns related to business and residential areas.
• Demonstrated reduced computational requirements suitable for edge deployment.

Conclusions:

• HAC-ULECA offers an efficient method for optimizing BSS rebalancing operations.
• The algorithm provides valuable insights into urban mobility patterns through BSS data.
• Edge computing integration enhances the functionality of BSS as urban sensor platforms.