A Spatio-Temporal Entropy-based Framework for the Detection of Trajectories Similarity.

Area of Science:

• Data Science
• Urban Computing
• Geospatial Analysis

Background:

• The increasing volume of sensor data and big data repositories presents opportunities for data science.
• Analyzing large-scale urban trajectory datasets for movement patterns and outliers is an underexplored research area crucial for urban management and planning.

Purpose of the Study:

• To introduce a novel spatio-temporal framework, STE-SD (Spatio-Temporal Entropy for Similarity Detection).
• To quantitatively evaluate the spatial and temporal distribution of complementary trajectory descriptors.
• To apply the framework to identify patterns and outliers in urban trajectories.

Main Methods:

• Developed a spatio-temporal framework (STE-SD) extending Shannon's entropy concept to spatial and temporal dimensions.
• Considered and quantitatively evaluated distributions of trajectory primitives: curvatures, stop-points, self-intersections, and velocities.
• Utilized entropy for identifying and qualifying these primitives within urban trajectories.

Main Results:

• The STE-SD framework successfully analyzes urban trajectories by evaluating spatial and temporal distributions of movement primitives.
• Demonstrated the framework's capability in identifying patterns and potential outliers within the Geolife dataset from Beijing.

Conclusions:

• The proposed STE-SD framework offers a robust method for uncovering insights from large urban trajectory datasets.
• This approach enhances the understanding of human movement in cities, supporting improved urban planning and management strategies.