Abstract and the results of the classification are reported;

Abstract

From the last few years
artificial neural network is playing a very important role in business
analytics and applications.  On studying
the application of artificial neural network in the field of marketing and
business it revealed that most of the work is done on the financial distress
and bankruptcy problems, stock price forecasting, and decision support, with
special attention to classification tasks. Also the application of artificial
neural network  in the market
segmentation. In this ANN in segmentation is analyzed and the results of the
classification are reported; and finally, the conclusions, limitations and
implications of the study are discussed.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

Introduction

Artificial neural
network are the computing systems based on the biological neural networks that
constitutes the human brain. It can be explain well with an example of image
recognition, to recognize a cat they do it by using their prior knowledge about
the cats, i.e. they have furs, tails, whiskers, cat like face etc. The
characteristics of artificial neural networks such as efficiency, robustness
and adaptability make them a valuable tool for classification, decision
support, financial analysis or credit scoring made its utilization in various
fields for example scientific fields as well as in many business applications.
Wong has reviewd the papers and articles published in 1988-1995 to study the
role of artificial neural network in business. But he has seen that most of the  research was done in the bankruptcy
predictions and stock forecasting. Later he also studied the work of it in the
collection of data and analyzing of these data. 
There are also various disciplines that have been studied including
accounting, costs monitoring, customer analysis, finance, marketing or sales,
manufacturing, process optimization, engineering or operational research have
not been included. In the second paper we have studied the application of ANN
in segmentation.

Research
Methodology

To study this we have
use the keywords “literature review” “artificial neural network in marketing””business”,
“finance”, “corporate”, “stocks”, “capital”, “costs”, “financial analysis”,
“accounting”, “bankruptcy”, “exchange rates”, “financial distress”,
“inflation”, “marketing”, “customers”, and “bonds”.

It is essential to say
that maximum of the information gather from the articles were studied and
utilized.

In the second
paper to study the application of ANN in segmentation Expert systems (ES) with

application and
Information system are the most common approach.

 

Application
area

Neural networks
captures data by using itretive algorithms by comparing there synaptic weights.
But the main disadvantage was that it considers only the data with large
weights and do not consider data with small data because small data do not
provide significant result. Primarily due to unavailability of data researchers
use artificial data. Application of neural network in the field of business is
very significant because it is use to extract valuable information from
complex, nonlinear and noisy data. The applications of neural network in
business are as follows:

·        
Auditing and accounting

·        
Cost monitoring

·        
Credit scoring

·        
Customers metrics

·        
Decision support

·        
Derivatives

·        
Exchange and interest rates

·        
Financial analysis

·        
Financial distress and bankruptcy

·        
Fraud analysis

·        
Inflation

·        
Marketing

·        
 Sales

·        
Shares and bonds

In the second
paper we have studied that in market segmentation methods can be largely
classified based on two criteria for the four categories: a priori or post hoc,
and descriptive or predictive statistical methods.When the type and number of
segments are determined in advance by the researcher then the apriori approach
is used and when the type and number of segments are determined based on the
results of data analyses then the post hoc approach is used. The post-hoc
methods are relatively powerful and frequently used in practice . A single set
of segmentation bases that has no distinction between dependent and independent
variables are related with the descriptive methods. When one set consists of
dependent variables to be explained or predicted by a set of independent
variables then the predictive methods are applied.

There are four
major classes of traditional algorithms for conducting traditional post hoc
segmentation studies: Cluster analysis,Correspondence analysis, Search
procedures, and Q-type factor analysis. Among clustering methods, the K-means
method is the most frequently used. An unsupervised neural network of the
artificial neural networks (ANNs) where the

outcomes are not
a priori have been recently applied to a wide variety of business areas. The

Kohonan
Self-Organizing Map of unsupervised ANN used in clustering for large and
complex

data.

 

Neural networks

In
the application of neural networks in business almost all types of neural
networks are used. But there are cases in which uncertain work on neural
network is done. So there is additional work should be done on these neural
network so that we can get outcomes.

 

Types of neural network

The
most popular neural networks used in the study was multilayer feedforward
neural networks in which neurons are organized into series of layers and
information signal flows through the network solely in one direction, from the
input layer to the output layer.

Classification of framework

Cluster
analysis is a common tool for market segmentation. Conventional research
usually employs the multivariate analysis procedures. Comparison of three
clustering methods were done and proposed that SOM performs better clustering
than the other conventional methods. A data mining associatiation ruile based
on SOM has been developed and applied to a sample of sales records from
database for market fragmentation. It was found that NN models outperforms the
multinomial logut model in determining the most profitable time in a purchasing
history to classify and target prospective consumers new to their categorie.
deployed an ANN guided by genetic algorithms (GAs) successfully to target
households. Targeting of customer segments withtailored   promotional activities is an important
aspect of customer relationship management. Application of  the SOM networks to a consumer data set the
research established that the SOM network performs better than the two-step
procedure that combines factor analysis and K-means cluster analysis in
uncovering market segments.All the selected articles were individually reviewed
and categorized based on the proposed classification framework by the authors
of this paper. The proposed classification scheme is consisting of the
following phases:

·        
Online
database search

·        
Initial
classification by the researcher

·        
Verification of the classification
result

Classification of the articles

The
following 14 types of ANN algorithms are found to be applied on market
segmentation research from the year 2000 to 2010 in the selected reviewed journals:
i) NN algorithm, ii) Meta Heuristic tools, iii) ARNN(association reasoning
neural network), iv) ART2 v) Bayesian NN, vi) Back Propagation NN, vii) Data
Mining, viii)hybrid fuzzy tools, ix) Genetic Algorithm(GA), x) hopefield NN,
xi) hybrid NN, xii) SelfOrganizing Map(SOM), xiii) support vector machine
(SVM), xiv) Vector Quantization.

 

Learning algorithm

The
process by which neural network updates its free parameters to capture the
patterns in the presented sample is called the learning. The most common
algorithm used  in reviewing business
applications was the backpropagation learning performed by gradient descent
search. This method is generally used because of its simplicity, universality
and good availability in softwares.

 

Hybridization

The
group of hybrid networks may be divided into two categories depending upon the
methodology used: (a) dealing with learning process, (b) dealing with net-work
architecture. The use of hybrid neural network is always having more importance
than ordinary neural networks.

 

Benchmark method

By
using neural network method we can get better results rather than the
conventional method. The most common benchmark methods identified in our
research are discriminant analysis , linear regression , logit and ARIMA. The
significant advantage of using conventional methods is their transperancy and
capability to comprehensibly interpret received results.

 

Citations

Number
of citations contains information about which researcher is interested in.  

Even
though the probability of being cited depends on various factors such
aspublication time, journal accessibility, or field, citation count is
anattractive measure for the evaluation of scientific performance.

 

Journal

The
review paper surveyed about a total number of 125 identified journals, first
six journal have published 201 (53.60%) papers and obtained 16,428 (58.71%)
citations. Large number of involved journals indicates that the contribution of
neural networks is scattered across a wide range of different business
applications.

 Most of the journals revealed that most of the
neural networks considered only real world application but not the underlying
factors such as economic and financial theory.

 

 

Conclusions

In
last few decades artificial neural network has progressed very much. It has
various applications in business fields but there were so much less papers were
published in this field.  In our study
were financial distress and bankruptcy analysis, stock price prediction, and
credit scoring. It is interesting that the average number of financial analysis
and derivatives articles stayed approximately the same throughout the examined
period. On the other hand, research on shares, marketing, financial distress,
and credit scoring has significantly increased compared to the early years of
our survey. After using neural networks in the business fields there were also
the fields which were not investigated. This is true not only for the
qualitative data but it also includes the quantative data just like cost, debt
financing and bonds.

In
hybridization secondary methods perform much better than the traditional
feedforward networks trained by gradient based techniques. The specific hybrid
networks might work well only for particular tasks, our survey suggests that
proper integration of met heuristic methods into the neural network methodology
might be a key for achieving the optimal performance.

Neural
networks have been successfully applied in wide range of business tasks and
were able to detect complex and nonlinear relationships without requiring any
specific assumptions about the distribution or characteristics of the data.
There is  lack of formal background and
the explanatory abilities are the two essential problems that have to be
resolved to improve the neural network business studies. The further research
therefore should focus on universal guidelines and general methodology for the
setting of control variables, selection of hidden layers and overall design of
the topology, since the quality of models reviewed in this study considerably
depended on experiences of the researchers. Moreover, robust measures that
could assess the relevance of individual explanatory variables are very
desirable, since researchers are currently still careful with interpretation of
their results and perform their validation using conventional methods. We are
convinced that research on artificial neural networks in business has still
much to offer. With their undisputed advantages, general availability of data
and increasing user-friendliness of soft-ware packages, neural networks will
surely attract more authors and offer additional possibilities for
applications.

Application
of artificial neural network techniques in market segmentation is an emerging
inclination in the industry and academia. It has paying the attention of
researchers, industry practitioners and academics. This work has identified
sixty four articles related to application of artificial neural network
techniques in market segmentation, and published between 2000 and 2010. This
article aims to give a research review on the application of neural network in
the market segmentation domain and techniques which are most often used. While
this review work cannot claim to be exhaustive, but it presents reasonable insights
and shows the prevalence of

research
on this area under discussion.

·        
The majority of the reviewed
articles  34.38% (22articles) and 21.88%
(14 articles) are related common neural network algorithms.

·        
Thus a trend of ANN research to
segmentation is more obvious from the articles published in the kind of journal
related to expert system development.

·        
These articles could provide insight to
organization strategists on the familiar artificial neural network practices
used in market segmentation.

·        
There are relatively fewer articles with
the metaheuristic, ART2, data mining, Genetic Algorithm and fuzzy algorithms.
Despite the fewer number of articles related to the above category of
artificial neural network application to market segmentation, it does not mean
the application of artificial neural network in this aspect is less mature than
in the others. Applications of those algorithms in other domains, such as
clustering or classification, may also be applied in segmentation if they
possess the same purpose of analysing the distinctiveness of customers/market.

·        
The k-means clustering model is the most
commonly applied model in segmentation by partitioning a large market into the
smaller groups or the clusters of customers.

·        
In order to maximize an organization’s
profits through segmentation, strategists have to both segment the market and
thus increase the profitability of the organisation.