When Traffic Flow Prediction Meets Wireless Big Data Analytics

Yuanfang Chen, Mohsen Guizani, Yan Zhang, Lei Wang, Noël Crespi, Gyu Myoung LeearXiv 2017 · arXiv

This paper, published in 2017, explores the intersection of traditional traffic engineering and the modern "Big Data" era. It provides a roadmap for how wireless data—generated by everything from GPS units to roadside sensors—can be used to solve one of the most expensive problems in modern cities: traffic congestion.

Introduction

Traffic congestion is a global crisis, costing the U.S. alone an estimated $160 billion annually in wasted fuel and time, while contributing heavily to CO2 emissions and public health risks. The core challenge is prediction. If we can accurately predict traffic flow, we can implement "eco-friendly routing" and better city planning. However, traffic flow is complex; it is a "continuous" phenomenon where the state of one road depends on the roads feeding into it. Traditional prediction models often struggle to handle the sheer volume, variety, and velocity of data generated by modern Intelligent Transportation Systems (ITS).

The authors demonstrate that traffic flow prediction can be significantly improved by using wireless "Big Data" to dynamically optimize the parameters of traditional time-series models. By fitting an ARIMA model to real-world trajectory data, the authors achieved a nearly 12-fold reduction in prediction error compared to non-optimized models.

A Framework for Data-Driven ITS

The authors propose a general framework for transforming raw wireless signals into actionable urban intelligence. This architecture moves from data collection to high-level applications through a series of analytical layers.

The proposed hierarchical framework for traffic prediction and management.
  1. Data Collection: Harvesting raw wireless signals (GPS, GSM, WiFi, Bluetooth, RFID).
  2. Trajectories (Spatio-Temporal Data): Extracting features like time sequences, locations, and geographical information.
  3. Knowledge (Prediction): Using Deep Learning, Markov Chains, or optimized time-series models to analyze and mine the data.
  4. Application: Turning that knowledge into city planning, transportation controlling, and management.

Methodology and Dataset

The study utilizes a robust real-world dataset from the England transportation system, covering 2,501 traffic roads across 300 miles of highways and arterial roads. Data is updated every 15 minutes, providing high-resolution vehicle trajectory sequences.

The authors focus on optimizing the ARIMA (Autoregressive Integrated Moving Average) model. The performance of ARIMA depends on three parameters: $p$ (autoregressive terms), $d$ (differences), and $q$ (moving average terms). Instead of using static parameters, the authors use the Bayesian Information Criterion (BIC) to analyze massive amounts of historical trajectory data and select the optimal $p, d,$ and $q$ values for specific road segments.

Key Results

The experimental results were striking. By comparing a "Normal ARIMA" model against their "Optimized ARIMA" (tuned via BIC and big data), they found:

  • Drastic Error Reduction: In one case study (March 15, 2015), the Root-Mean-Square Error (RMSE) dropped from 64.05 in the normal model to just 5.38 in the optimized model.
  • Consistency: Across a 31-day average in March 2015, the optimized model consistently outperformed the standard version, proving that the performance gain wasn't a fluke of a specific day's traffic.
  • Trend Tracking: The optimized model was far better at capturing the "peaks" of rush hour, which is the most critical time for congestion management.
Comparison of actual traffic flow vs. normal and optimized ARIMA models across different dates and monthly averages.

Limitations and Future Work

While the results are promising, several hurdles remain:

  • Data Integrity: Wireless sensors are prone to failure. If a critical sensor goes offline, the "Big Data" becomes "Missing Data," which can lead to dangerously incorrect predictions.
  • Privacy Protection: Trajectory data contains sensitive information (locations, plate numbers). Finding a way to analyze this data without compromising individual privacy is a major research challenge.
  • Complexity of Mobility: While ARIMA works well for flow, it doesn't fully capture complex human behaviors. Future work should integrate more Deep Learning and hybrid models to handle these non-linear patterns.

Summary updated

Abstract

Traffic flow prediction is an important research issue for solving the traffic congestion problem in an Intelligent Transportation System (ITS). Traffic congestion is one of the most serious problems in a city, which can be predicted in advance by analyzing traffic flow patterns. Such prediction is possible by analyzing the real-time transportation data from correlative roads and vehicles. This article first gives a brief introduction to the transportation data, and surveys the state-of-the-art prediction methods. Then, we verify whether or not the prediction performance is able to be improved by fitting actual data to optimize the parameters of the prediction model which is used to predict the traffic flow. Such verification is conducted by comparing the optimized time series prediction model with the normal time series prediction model. This means that in the era of big data, accurate use of the data becomes the focus of studying the traffic flow prediction to solve the congestion problem. Finally, experimental results of a case study are provided to verify the existence of such performance improvement, while the research challenges of this data-analytics-based prediction are presented and discussed.
BibTeX
@misc{chen2017when,
  author = {Yuanfang Chen and Mohsen Guizani and Yan Zhang and Lei Wang and Noël Crespi and Gyu Myoung Lee},
  title = {{When Traffic Flow Prediction Meets Wireless Big Data Analytics}},
  year = {2017},
  eprint = {1709.08024},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1709.08024},
}
Figure