Recommendation for Ridesharing Groups Through Destination Prediction on Trajectory Data

Tang, Lei; Duan, Zongtao; Zhu, Yishui; Ma, Junchi; Liu, Zihang · 2019 · OpenAlex-citations

DOI: 10.1109/tits.2019.2961170

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of optimizing passenger matching in ridesharing systems, specifically focusing on "OD-slugging," where passengers share rides with drivers whose routes do not strictly align with their origins or destinations. Existing methods often rely on manually entered origin-destination coordinates, which ignores spatial semantics, social interactions, and the inaccuracies inherent in user-inputted addresses, particularly in unfamiliar areas. The authors aim to improve matching accuracy and reduce total travel time by predicting passenger destinations from historical GPS trajectory data and grouping passengers based on these predicted destinations and behavioral markers. The proposed method consists of two main components: a destination prediction algorithm and a group recommendation optimization model. For destination prediction, the authors developed a PrefixSpan-prediction using partial matching (P-PPM) algorithm. This approach mines frequent movement patterns from semantic trajectory data—sequences of locations enriched with type information (e.g., science, education)—to determine the confidence of movement rules. The P-PPM model utilizes historical departure times, coordinates, and location types to predict the probability distribution of a passenger’s intended destination. For group recommendation, the system identifies potential passengers near a driver’s origin within specific spatial and temporal constraints. It then formulates an optimization problem to minimize the total travel time of the group, accounting for extra walking distances to pick-up points and additional waiting times. The algorithm selects a group of passengers that fits within the vehicle’s capacity and minimizes the combined travel time, including the driver’s route and passengers’ walking segments. Experimental results demonstrate that the proposed method significantly outperforms baseline approaches. The accuracy of destination prediction increased from 46% to 80% compared to existing methods. Furthermore, the ridesharing scheme achieved substantial efficiency gains, with passengers saving over 19% of total travel miles. The system effectively identified groups of strangers with similar travel preferences rather than relying on social connections, thereby expanding the pool of potential matches. The optimization model successfully balanced the trade-offs between walking distance, waiting time, and vehicle capacity, providing feasible and efficient ridesharing arrangements. The significance of this work lies in its ability to enhance the user experience and efficiency of ridesharing services by automating destination prediction and group formation. By leveraging trajectory data and semantic information, the method reduces the burden on users to manually input precise destinations and improves the accuracy of matching. This approach supports more flexible and scalable ridesharing systems, particularly in urban environments where traffic congestion and parking costs are high. The findings suggest that integrating destination prediction with group optimization can lead to more effective utilization of shared mobility resources, benefiting both passengers and drivers.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction