Directed-edge-based mining of regular routes for enhanced traffic pattern recognition from travel trajectories

Trajectory analysis serves as a critical technique for uncovering patterns in users' transportation behaviors. This paper introduces a direction-aware regular route mining algorithm that systematically processes GPS (Global Positioning System, GPS) trajectory data by preprocessing, detecting stay regions, segmenting abnormal trajectories, and extracting frequent directed edges and supporting paths through route clustering, ultimately constructing users' regular travel routes. By integrating stop-rate features, the algorithm effectively distinguishes between transportation modes, such as public transit and private vehicles. Experimental results based on the Geolife dataset, which includes trajectory data from 208 users covering a total distance of 1.35 million kilometers, indicate that the proposed method reduces the Mean Absolute Percentage Error (MAPE) by 56%, 49%, and 32% compared to the Rules-based method, CNN (Convolutional Neural Network, CNN), and DBSCAN (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) algorithms, respectively, in both regular route extraction and transportation mode recognition. This improvement highlights the algorithm's enhanced accuracy in identifying travel patterns. The proposed approach offers valuable support for applications in dynamic traffic prediction and personalized route recommendation systems.