COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(10) 324-328 Sun Li-ping, Luo Yong-long, Ding Xin-tao, Chen Fu-long
324
Operation Research and Decision Making
Spatial clustering algorithm with obstacles constraints
based on artificial bee colony
Li-ping Sun
1, 2
, Yong-long Luo
1, 2
, Xin-tao Ding
2
, Fu-long Chen
2
1
College of National Territorial Resources and Tourism, Anhui Normal University, Anhui, 241000, China
2
Engineering Technology Research Center of Network and Information Security, Anhui Normal University, Anhui, 241000, China
Received 1 August 2014, www.cmnt.lv
Abstract
Spatial clustering is one of practical data mining technique. In this paper, artificial bee colony (ABC) is used for clustering algorithm,
which aims to optimally partition N objects into K clusters in obstacle space. The ABC algorithm used for clustering analysis with
obstacles constraints, called The ABC algorithm used for clustering analysis with obstacles constraints ABC-CO, is proposed in the
paper. By comparison with the two classic clustering algorithms, k-medoids and COE-CLARANS, demonstrates the rationality and
usability of the ABC-CO algorithm.
Keywords: spatial clustering, artificial bee colony, obstructed distance; fitness calculation
1 Introduction
Spatial clustering is the organization of geographical data
set into homogenous groups, the aim of which is to group
spatial data points into clusters [1-3]. Most spatial
clustering algorithms apply Euclidean distance between
two sample points to measure the proximity of spatial
points. However, physical obstacles (e.g. rivers and
highways) often exist in real applications, which can
hinder straight reachability among sample points. As a
result, the clustering results, which utilize Euclidean
distance measure are often unreasonable. Taking the
simulated dataset in Figure 1a as an example, where the
points represent the location of consumers. The clustering
result shown in Figure 1b can be obtained, when the rivers
and hill as obstacles are not considered. Obviously, the
result is not realistic. If the obstacles are taken into
account, the clustering result in Figure 1c can be obtained.
a) Spatial dataset with obstacles
b) Spatial clustering result ignoring obstacles
c) Spatial clustering result considering
obstacles
FIGURE 1 Spatial clustering in the presence of obstacles
At present, there are a few algorithms considering
obstacles constraints in the spatial clustering process [4-9].
There generally exists the shortcomings, including low
robustness and easy to fall into local optimum. Many
heuristic clustering algorithms have been introduced to
overcome local optima problem, such as evolutionary
algorithms [10, 11], swarm intelligence algorithms [12,
13] and so on [14, 15]. Artificial bee colony (ABC)
algorithm, which simulates the intelligent foraging
behaviour of a honey bee swarm, is a novel category of
heuristic algorithms. In this paper, ABC optimization
Corresponding author e-mail: ylluo@ustc.edu.cn
algorithm for optimization problems, which is proposed by
Karaboga [16], is applied to spatial clustering analysis.
Fathian and et al. proposed the application of honeybee
mating optimization in clustering [17]. Zhang and et al.
presented a novel artificial bee colony approach for
clustering which was compared with other heuristic
algorithms such as genetic algorithm, ant colony,
simulated annealing and tabu search [18]. Yan and et al.
designed a hybrid artificial bee colony (HABC) algorithm
for data clustering [19]. Karaboga and Ozturk applied