An Intelligent System for Taxi Service Monitoring, Analytics and Visualization

Yu Lu, Gim Guan Chua, Huayu Wu, Clement Shi Qi OngIJCAI 2016

Introduction

In dense urban environments, the balance between taxi supply and passenger demand is a persistent challenge. This "spatio-temporal imbalance" often results in passengers waiting indefinitely for a ride while drivers circle empty or wait in inefficient queues. While modern taxi fleets generate massive amounts of GPS and telematics data, converting this "firehose" into actionable intelligence for fleet operators and city planners remains a significant technical hurdle. This paper presents an intelligent system designed to monitor, analyze, and visualize citywide taxi services in real-time.

System Architecture

The proposed system bridges the gap between raw data collection and high-level decision support through a modular architecture. It processes both historical data for long-term pattern recognition and buffered streaming data for real-time updates.

The system architecture showing the data flow and the three core analytical modules.

As illustrated in the architecture, the system is divided into three primary analytical components:

  1. Wait Time Estimation Module: This acts as the system's "brain," calculating the expected durations for both passengers to find a taxi and drivers to secure a fare.
  2. Hotspot Detection Module: Identifies high-demand areas using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm on historical state-transition data (e.g., where a taxi switches from "FREE" to "HIRED").
  3. Trip Extraction Module: Reconstructs complete passenger journeys by stitching together GPS pings from the start of a trip (POB) to the final payment.

Analytical Innovations

A core challenge in taxi analytics is the "unobservable" nature of street-hail demand; systems typically know where taxis are, but not necessarily where passengers are waiting. To solve this, the researchers introduced the Booking Ratio—the percentage of trips initiated via apps or calls versus street hails. A high booking ratio in a specific region serves as a proxy for severe street-hail wait times, as passengers opt for paid bookings when cabs are scarce.

Furthermore, rather than attempting to predict wait times down to the exact second—which is often noisy and inaccurate—the system outputs four practical categorical levels: Short (<2 min), Reasonable (2-5 min), Large (5-10 min), and Severe (>10 min).

Visualization and Results

The system was validated using data from approximately 26,000 taxis in Singapore, collected via Mobile Data Terminals (MDTs) and transmitted via 3G/GPRS. The results are presented through a sophisticated 3D Visualization Engine that maps wait times to color-coded regions and displays trip flows as dynamic arcs.

3D visualization of (a) passenger wait times, (b) taxi wait times, (c) trip flows, and (d) detected hotspots.

The dashboard provides several key operational insights:

  • Wait Time Mapping: Color-coded regions (as seen in Figure 4a) highlight "red zones" where wait times exceed 10 minutes, typically in remote or underserved sectors.
  • Queue Dynamics: At major hubs like Changi Airport, the system can visualize the physical length of taxi queues and estimate the "burn rate" of how fast those taxis are being dispatched.
  • Flow Visualization: Dynamic arcs (Figure 4c) represent the directional movement of taxis, showing the real-time "drain" of vehicles from one part of the city to another.

Conclusion and Future Work

The system successfully transforms raw telematics into an intuitive operational dashboard. While current capabilities focus on real-time monitoring, future work aims to incorporate predictive forecasting to anticipate demand 30 minutes in advance, allowing drivers to move toward demand before it peaks. Additionally, integrating external factors like weather data or public transit disruptions could further refine the accuracy of wait-time estimations, as rain and train delays significantly influence urban taxi demand.

Summary updated

BibTeX
@inproceedings{lu2016intelligent,
  author = {Yu Lu and Gim Guan Chua and Huayu Wu and Clement Shi Qi Ong},
  title = {{An Intelligent System for Taxi Service Monitoring, Analytics and Visualization}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2016},
}
Figure