An Adaptive Trajectory Clustering Method Based on Grid and Density in Mobile Pattern Analysis.

Area of Science:

• Data Mining
• Computational Geometry
• Geographic Information Systems

Background:

• Trajectory clustering is crucial for applications like hotspot detection and urban mobility analysis.
• Existing methods require manual parameter tuning, increasing workload and complexity.
• Adaptive parameter calibration is needed to streamline trajectory clustering.

Purpose of the Study:

• To propose an adaptive trajectory clustering approach (ATCGD) that automates parameter calibration.
• To reduce the computational complexity and workload associated with trajectory clustering.
• To improve the efficiency of trajectory data mining.

Main Methods:

• The ATCGD approach involves three stages: partition, mapping, and clustering.
• Segments are partitioned using the average angular difference-based MDL (AD-MDL) method.
• Clustering utilizes a DBSCAN-based method with parameters adapted from the mapping stage.

Main Results:

• ATCGD significantly reduces clustering runtime by approximately 95% compared to TRACLUS.
• Adaptive parameter calibration shows a difference of less than 5% compared to optimal values in most cases.
• The approach maintains high partition accuracy while decreasing segment count.

Conclusions:

• The proposed ATCGD method offers an efficient solution for trajectory clustering.
• Automated parameter calibration in ATCGD reduces workload and computational cost.
• ATCGD demonstrates a practical advancement in trajectory data mining applications.