Repeatability and Similarity of Freeway Traffic Flow
and Long-Term Prediction Under Big Data
Zhongsheng Hou, Senior Member, IEEE, and Xingyi Li
Abstract—In this paper, by splitting a traffic flow series into
basis series and deviation series, the concepts of similarity and
repeatability of traffic flow patterns are defined using the statistic
average values of the basis series and the deviation series and are
further verified through the real-time big traffic data of 82 days
with a sampling period of 5 min collected from two typical ones
among a total of 102 detecting sites in Shenzhen, China. Mean-
while, based on the repeatability and the similarity of the traffic
flow series, a novel long-term forecasting method for traffic flow is
developed, and hybrid forecasting algorithms for short-/long-term
traffic flow prediction are also proposed. The effectiveness of these
algorithms is verified by using the real-time data.
Index Terms—Big data, freeway traffic flow, long-term predic-
tion, repeatability, similarity, short-term prediction.
I. INTRODUCTION
T
HE congestion of the freeway traffic caused by heavy
traffic demand in many Chinese metropolitans, such as
Beijing, Shanghai, Shenzhen, is becoming a routine pheno-
menon, which leads to a large uncertainty for the drivers using
freeways. However, with the development of advanced sensor
techniques, such as inductive loops, radar sensors, and video
cameras, the real-time traffic big data can now be collected
more easily. In the sequel, the traffic patterns may be discov-
ered and exploited if these big data can be effectively processed
and analyzed. Using the real-time traffic big data collected and
the patterns discovered, much useful information such as real-
time freeway trafficstate, travel time and routeinformation,can
be provided to the traffic authority and freeway users. In such a
circumstance, the traffic administrator may supply the freeway
informationthroughvariablemessagesigns (VMS)or on-board
navigators to the drivers for choosing their travel mode, route,
starting time or even canceling their trips.
Manuscript. received October 15, 2014; revised May 27, 2015, November 12,
2015, and November 23, 2015; accepted December 13, 2015. Date of publica-
tion January 26, 2016; date of current version May 26, 2016. This work was
supported in part by the State Key Project Program of the National Natural
Science Foundation of China under Grant 61433002 and in part by the Major
Program of International Cooperation and Exchanges of the National Natural
Science Foundation of China under Grant 61120106009. The Associate Editor
for this paper was W.-H. Lin.
Z. Hou is with the Advanced Control Systems Laboratory, School of
Electronic and Information Engineering, Beijing Jiaotong University, Beijing
100044, China (e-mail: ).
X. Li is with the School of Computer Science and Telecommunications
Engineering, Jiangsu University, Zhenjiang 212013, China (e-mail: lixinyii@
163.com).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TITS.2015.2511156
Traffic flow prediction (TFP) is an important issue for both
the transport authority and the freeway users. Using the TFP
results,trafficadministratorscan carryoutmorepowerfultraffic
managementand apply more effective control measurements to
weaken the traffic congestion and to promote efficiency of the
freeway networks. Using the real-time TFP results, the freeway
users can make their decisions on travelling modes or choosing
routes to reach their destinations. TFP, as the basis of traffic
planning and management, is one of the most important re-
search topics in the field of traffic sciences. Generally, TFP can
be divided into three classes, short-term forecasting, medium-
term forecasting and long-term forecasting. The one with the
time span from 5 minutes to 30 minutes is usually regarded as
the short-term forecast, that from 30 minutes to a few hours
is the medium-term forecast, and that from one day to several
days is the long-term forecast.
The study of the short-term forecast of traffic flow attracts
ever increasing attentions and has gotten lots of developments,
and many results have been applied in practice [1]–[3]. These
methods may be cataloged by using the different mathematics
theories, such as state-space methods [7], Kalman filter meth-
ods [4], [5], spectral analysis methods [6], statistical tech-
niques [8]–[10], regression models [11]–[13], ARIMA models
[14]–[16], [40], Bayesian dynamic linear model approach [17],
Markov model [18], non-parametric regression method [16],
[19]–[21], chaos techniques [22], fuzzy techniques [23], neural
networks [24]–[26], support vector machine [27], genetic al-
gorithm [28], linear genetic programming [29], grey system
model [30], data driven methods [31], and others [32]. These
methods reveal many intrinsic properties of traffic flow from
different mathematical angles. On the other hand, they could
also be divided into two big catalogues. One is the model-
based method, and the other is the data driven method. It is
worth pointingoutthat, accuratepredictionoftrafficflow is still
very difficult to acquire due to the existence of many external
disturbance factors, and thus the reliable model-based or data-
based traffic flow prediction is always a hot topic of transporta-
tion research.