17-03-2014, 10:05 PM
Abstract—Grade on the major hazards is of great
importance to industry. But the already proposed
methods are not fit in with the precision we need. In this
paper, a novel method is proposed for dynamic grade on
Major Hazards using Community Detection in complex
networks (namely MHCD). Firstly MHCD represents the
input data as a network, and then uses a novel
evolutionary algorithm to find the communities in such a
network. Each detected community corresponds to a
specific risk grade. In this work we introduce a new
generalized method for transformation of the input data
to network, and propose a novel evolutionary algorithm
to detect the communities. The results of the simulation
experiment on a practical problem show that compared
with other classification methods, MHCD has better
performances.
Keywords-major hazards grading; complex network;
community detection; genetic algorithm
I. INTRODUCTION
Major hazards in industry will bring great disasters
if the state of its substances or energies is destroyed,
which will bring great losses not only to plants and
mines, but to the personnel. It has aroused the
attractions of the people from all over the world. It is
placed in the first place that to differ the grades of the
major hazards if one want to take effective measures
according to the grades of major hazards [1].
Grade on major hazards is usually done in term of
scoring, which based on fixed grades and is done once
for all. When one hazard is categorized into a specific
grade with regard to certain standards, the grade of
such a hazard will not be changed. The scores,
however, will be different from person to person
because of subjective factors. Worse still, the number
of the objects to be graded is very large, it is
impossible to directly grade on the total objects. So the
much more reasonable way is to sample on the total
objects, and based on the sampled objects to establish a
standard for grade. So the grade can not be finished
once for all, and should be modified with the
introduction of new objects. So the grade on major
hazards should be dynamic [1].
Evolutionary algorithm [2] [3] [4] (EA) is firstly put
forward by Holland and his colleagues. It, based on
Darwin’s theory of the survival of the fittest and
Mendel’s theory of heredity, is one of the populationbased
approaches, and, with many good features such
as intelligence, concurrence, and robustness. It requires
little problem knowledge to set up an optimization
problem, and that individuals are created through semirandom
or random operators, such as mutation and/or
recombination, that are easy to implement. It is a most
effective method, which can be used to solve real
world problems with high complexity [5] [6] [7]. EA
starts from a so-called population, a group of
individuals, and use strategies like crossover, mutation
and selection to lead the population to a better direction
to get global optimum. So EAs are naturally fit in with
the problems about the grade on major hazards. In this
paper, a novel method for dynamic grade on Major
Hazards using Community Detection in complex
networks (namely MHCD) is proposed. It represents
the input data as a network, and uses a novel
evolutionary algorithm to detect the communities in
such a network. Each detected community corresponds
to a specific risk grade. In this work we introduce a
new generalized method for transformation of the input
data to network, and propose a novel evolutionary
algorithm to detect the communities. The results of the
simulation experiment on a practical problem show that
compared with other classification methods, MHCD
has better performances.
importance to industry. But the already proposed
methods are not fit in with the precision we need. In this
paper, a novel method is proposed for dynamic grade on
Major Hazards using Community Detection in complex
networks (namely MHCD). Firstly MHCD represents the
input data as a network, and then uses a novel
evolutionary algorithm to find the communities in such a
network. Each detected community corresponds to a
specific risk grade. In this work we introduce a new
generalized method for transformation of the input data
to network, and propose a novel evolutionary algorithm
to detect the communities. The results of the simulation
experiment on a practical problem show that compared
with other classification methods, MHCD has better
performances.
Keywords-major hazards grading; complex network;
community detection; genetic algorithm
I. INTRODUCTION
Major hazards in industry will bring great disasters
if the state of its substances or energies is destroyed,
which will bring great losses not only to plants and
mines, but to the personnel. It has aroused the
attractions of the people from all over the world. It is
placed in the first place that to differ the grades of the
major hazards if one want to take effective measures
according to the grades of major hazards [1].
Grade on major hazards is usually done in term of
scoring, which based on fixed grades and is done once
for all. When one hazard is categorized into a specific
grade with regard to certain standards, the grade of
such a hazard will not be changed. The scores,
however, will be different from person to person
because of subjective factors. Worse still, the number
of the objects to be graded is very large, it is
impossible to directly grade on the total objects. So the
much more reasonable way is to sample on the total
objects, and based on the sampled objects to establish a
standard for grade. So the grade can not be finished
once for all, and should be modified with the
introduction of new objects. So the grade on major
hazards should be dynamic [1].
Evolutionary algorithm [2] [3] [4] (EA) is firstly put
forward by Holland and his colleagues. It, based on
Darwin’s theory of the survival of the fittest and
Mendel’s theory of heredity, is one of the populationbased
approaches, and, with many good features such
as intelligence, concurrence, and robustness. It requires
little problem knowledge to set up an optimization
problem, and that individuals are created through semirandom
or random operators, such as mutation and/or
recombination, that are easy to implement. It is a most
effective method, which can be used to solve real
world problems with high complexity [5] [6] [7]. EA
starts from a so-called population, a group of
individuals, and use strategies like crossover, mutation
and selection to lead the population to a better direction
to get global optimum. So EAs are naturally fit in with
the problems about the grade on major hazards. In this
paper, a novel method for dynamic grade on Major
Hazards using Community Detection in complex
networks (namely MHCD) is proposed. It represents
the input data as a network, and uses a novel
evolutionary algorithm to detect the communities in
such a network. Each detected community corresponds
to a specific risk grade. In this work we introduce a
new generalized method for transformation of the input
data to network, and propose a novel evolutionary
algorithm to detect the communities. The results of the
simulation experiment on a practical problem show that
compared with other classification methods, MHCD
has better performances.