2308 C. Luo et al.
(2) The crossover and mutation operators are controlled by adaptive probability functions.
Through adaptive adjustment of probabilities, the convergence precision of the hybrid
optimization algorithm is greatly improved. Thus, the convergence speed is accelerated.
(3) To assess the efficiency of the hybrid-LSSVM model, authentic traffic flow data are gath-
ered from the locations I80 corridor near Davis, CA. The experimental results show that,
compared to other predictive algorithms, the hybrid optimization algorithm has greater
prediction performance and computational efficiency for the short-term traffic flow pre-
diction. Therefore, the method proposed in this paper is more generally applicable.
The remainder of the paper is organized as follows. Section 2 defines the problem of
short-term traffic flow prediction and briefly describes the methodologies used in this study.
Section 3 details the proposed short-term traffic flow prediction method based on hybrid opti-
mization algorithm with LSSVM (hybrid-LSSVM). Section 4 presents the empirical analysis
and compares prediction results of several different prediction models. A brief conclusion of
this paper and main contributions are given in Section 5.
2 Background
In this section, we introduce some preliminary concepts of the main methodologies adopted
in this paper and define the problem of short-term traffic flow prediction.
2.1 Preliminary Concepts
Some preliminary concepts can be presented as follows.
(1) The PSO algorithm PSO is a typical heuristic algorithm, which is derived from the study
of flocking behavior of birds and insects. PSO solves a problem by generating a batch of
candidate solutions, here named particles. Every particle has its own position and velocity.
Each particle’s position indicates a potential solution to an optimization problem. Whether
the particle is excellent depends on its fitness value, which is calculated by a predefined
fitness function.
(2) The GA algorithm GA is a metaheuristic search algorithm, which is inspired by the
phenomena in evolutionary biology, including mutation, crossover and natural selec-
tion. GA solves a problem by generating initial candidate solutions randomly. Usually,
a solution can be expressed in the form of a chromosome, which comprises multiple
genes. According to the fitness value and selection strategies, favorable chromosomes
have higher probabilities to be selected to survive. Then, combining genetic operators
crossover and mutation to generate second generation solutions from the selected chro-
mosomes.
(3) Fitness function Fitness function is used to evaluate the performance of an optimization
method. Fitness value is a measure designed to represent the advantages and disadvan-
tages of an individual object in a problem by referring to the adaptability of an individual
to the environment. It is generally a real value function, which is the evaluation function
to guide the search for the optimal solution.
2.2 Problem Description
Traffic flow prediction refers to the use of historical, current traffic flow data and other relevant
statistical data to establish a suitable mathematical model. An intelligent calculation method
123