Monitoring ride-hailing passenger security risk: An approach using human geography data

Abstract

Ride-hailing services pose significant security challenges for passengers, underscoring the need for effective security risk monitoring. While extensive research has addressed various aspects of ride-hailing, few studies specifically focus on passenger security risk monitoring. This paper introduces onSecP, an online approach designed to monitor the security risks faced by ride-hailing passengers using human geography data. onSecP comprises two phases that set it apart from conventional anomalous trajectory detection methods. First, it employs an anomalous trajectory detection model using the LCSS-Kmeans-Geoinformation technique, which identifies and scores anomalous ride-hailing trajectories. Second, it utilizes a multi-parameter risk evaluation model enhanced by the AHP-Entropy-Cluster weighting method to perform real-time calculations of passenger security risks by integrating factors such as driver characteristics, trip details, geographical environment, trajectory anomaly scores, abnormal stop duration, and passenger information. Our approach leverages diverse data sources, including ride-hailing driver information, Point of Interest (POI) data as well as optimal route data from AMap, Global Positioning System (GPS) data, expert assessments, and passenger demographic surveys. Experimental evaluations demonstrate that onSecP effectively differentiates between unsafe trips and normal or abnormal trajectories, thereby significantly improving security risk monitoring for ride-hailing passengers. Consequently, onSecP offers a robust tool for enhancing ride-hailing security warning systems.