According to the similarity of data, the huge number of customers' data in e-supply chain can be clustered objectively and scientifically, and those customers are be clustered into corresponding groups based on SOFM ANN (Self-Organizing Feature Map Artificial Neural Network). Through recognizing and analyzing the different features of these different groups, adopting corresponding marketing strategies can enhance customers' satisfaction, and moreover, can realize the e-supply chain's benefit maximization. In this article, three aspects of improvement are made in the SOFM ANN that applied in customers' clustering analysis; the sample data comes from the Google group. The result shows that the improved SOFM ANN's performance is considerably better than the traditional one's performance. Customers' clustering analysis and corresponding marketing strategies based on the SOFM ANN is a comparatively new topic. Therefore, the result of the research in the article is only for reference.
Artificial Neural Network (ANN) is a new artificial in-
telligence technology. Since famous psychological dissec-
tor McCulloch and talent mathematician Pittst
[1]
have pro-
posed the 1st nerve model in 1943, and has experienced
1980s ANN development golden age, until today, the hu-
manity already had more than 60 years development histo-
ries to the ANN research. However, until now, the research
and application of ANN is still rising. ANN technology has
been widely used in areas such as pattern recognition, Sig-
nal Processing, Auto control, Fault Diagnosis, Communica-
tions, Electronics, Financial Forecasts, and Knowledge En-
gineering Management. Now, there are at least 30 kinds of
mature ANN models, well-known as BP neural networks,
Fukushima neural networks, Self-Organizing Feature Map
(SOFM ANN), Hopfield neural network, Boltzmann neural
networks, Elman neural networks, and so on. Among them,
SOFM ANN is of main concern to people and is used widely.
From the view of neurobiological basis, SOFM ANN
can be divided into two category of models. The first cate-
gory is proposed by Willshaw
[2]
that the dimension of input
is equal to the output ANN model, and the other is from Fin-
land Helsink professor at the University of Kohonem
[3]
of
the high-dimensional input mapping to the low-dimensional
output ANN model. As the latter does not lie in the details
of neural biology, but rather seizes the substantive charac-
teristics of calculate mapping in the human brain and re-
tains the ease of the calculation, the Kohonen model is more
widely applied. If unspecific, the SOFM ANN model usu-
ally refers to the Kohonen model. SOFM ANNs are more widely used in data compression and image recognition, but
application to customer relationship management, such as
clustering analysis and strategic studies in e-supply chain, is
a new research field.
In the clustering analysis and strategic studies based on
the SOFM ANN, according to the similarity of measured
customer data, we can classify the similar characteristics of
the customers into the same group using the characteristics
of the SOFM ANN. Realizing the identical group’s customer
difference minimum and the different group’s customer dif-
ference maximization, is advantageous for adopting the dif-
ferent marketing strategy to the different group’s customer
and adopting the similar marketing strategy to the identical
group’s customer. This finally realizes the e- supply chain
benefit maximization.
In the virtual Internet world of the e-supply chain, there
are magnanimous customer data. The clustering analysis and
strategic studies based on the SOFM ANN can effectively
reduce the huge management cost with fully personalized
marketing, and avoid the enormous risk of losing the 20%
potential VIP clients because of taking the same marketing
strategy for all customers. Thus, we acquire a Pareto opti-
mal equilibrium between the cost of customer service and
the profit of e-supply chain channel. In the traditional classi-
fication way for clustering analysis of the e-supply chain, as
the limitations of personal experience, we will artificially di-
vide the classification standards in prior, and it is difficult to
deal with the huge customer data with limitation of mental,
all of which can led to the emergence of errors in classify-
ing the customer groups, resulting in the wrong marketing
strategy. The customer data of e-supply chain are not only large and complex, and are usually non-structural or semi-
structured, and the traditional database technology can only
handle structured data; thus customer groups cannot be di-
vided scientifically. It is difficult to form a scientific cus-
tomer marketing strategy using the traditional database tech-
nology. In the e-supply chain customers’ clustering analysis
of this article, the ANN technology, the modern computer
technology is used to achieve a fully automated cluster group
according to the similarity of customer data. The objective of
this study is to make the complex nonlinear of customer data
packet processing more objective, scientific, and fair and rea-
sonable than the traditional methods.
Making clustering analysis on the customer behavior
characteristics in e-supply chain scientific and quick, for the
use of targeted marketing strategy, is crucial in ensure the
success of e-supply chain. Clustering analysis in the cus-
tomers of the e-supply chain based on the SOFM ANN has
naturally become a very interesting research topic. It is re-
grettable that despite the SOFM ANN can help people cap-
ture some surprised characteristic information of customers
from huge complex and unstructured data in the e-supply
chain, it is indeed very difficult to analyze the characteris-
tic of the SOFM ANN in mathematics at the general set.
Presently, people are able to provide some limited appli-
cation results, and unable to provide input data distribution
of the inherent credibility. There are also high-dimensional
mapping from the low-dimensional, prone to distortion, and
other issues. However, these do not affect the people’s in-
terests in research and applications of the SOFM ANN. Sev-
eral experts in the fields of ANN are committed to the im-
provement and application of the SOFM ANN clustering
algorithm. For example, people introduce the thought of
ant colony clustering thinking, genetic algorithm, and con-
science algorithm to improve the SOFM ANN clustering al-
gorithm, and have achieved fairly good application results.
In this article, an attempt has been made to apply the
improved SOFM ANN model into clustering analysis of cus-
tomers and the corresponding customers marketing strate-
gies in e-supply chain. The result and the method are only
for reference.
