Learning Fuzzy Network Using Sequence Bound Global Particle Swarm Optimizer

Learning Fuzzy Network Using Sequence Bound Global Particle Swarm Optimizer

Satchidananda Dehuri (Fakir Mohan University, India) and Sung-Bae Cho (Yonsei University, Korea)
Copyright: © 2012 |Pages: 17
DOI: 10.4018/ijfsa.2012010104
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

This paper proposes an algorithm for classification by learning fuzzy network with a sequence bound global particle swarm optimizer. The aim of this work can be achieved in two folded. Fold one provides an explicit mapping of an input features from original domain to fuzzy domain with a multiple fuzzy sets and the second fold discusses the novel sequence bound global particle swarm optimizer for evolution of optimal set of connection weights between hidden layer and output layer of the fuzzy network. The novel sequence bound global particle swarm optimizer can solve the problem of premature convergence when learning the fuzzy network plagued with many local optimal solutions. Unlike multi-layer perceptron with many hidden layers it has only single hidden layer. The output layer of this network contains one neuron. This network advocates a simple and understandable architecture for classification. The experimental studies show that the classification accuracy of the proposed algorithm is promising and superior to other alternatives such as multi-layer perceptron and radial basis function network.
Article Preview

Introduction

In the past few decades, information technology and the World Wide Web (WWW) have created stacks of innovations in the area of marketing style of companies. More businesses and companies are collecting highly informative and valuable data in a large scale. The huge amount of data can be a gold mine for business management and marketing. It is therefore increasingly important to analyze the data. However, timely and accurately processing tremendous volume of data with traditional methods (Michie et al., 1994) is a difficult task. For example, using multi-layer perceptron (MLP) in data mining is not likely produce any useful results (Edelstein, 1996), because it does not have a clean interpretation of the model and its longer training time can make frustration. The ability to analyze and utilize massive data lags far behind the capability of gathering and storing it. This gives rise to new challenges for businesses and researchers in the extraction of useful information.

Data mining-a core step of knowledge discovery in databases (Piatetsky-shapiro et al., 1996; Han & Kamber, 2001), is defined as a process of employing one or more computer learning techniques to automatically analyze and extract knowledge from vast amount of data contained within a database. The purpose of data mining is to identify trends and patterns in data. Classification (Duda et al., 2001) is one of the widely used techniques of data mining. In addition to classification, there are many important tasks such as association rule mining (Agrawal, 1993), clustering (Jains & Dubes, 1988), regression analysis (Chen, 2011), summarization, etc. in the area of data mining. However, classification is a fundamental activity in pattern recognition (Theodoridis & Koutroumbas, 2006; Bishop, 2006), data mining and so forth. Given predetermined disjoint target classes {C1, C2,...,Cn}, a set of input features {F1, F2, .., Fm} and a set of training data T with each instance taking the form <a1, a2, ..., am>, where ai <i=1,2,..., m> is in the domain of attribute Fi, i=1,2,…,m and associated with a unique target class label the task is to build a model that can be used to predict the target category for new unseen data given its input attribute values.

There are many classifiers like naïve Bayes (Robert, 2001; Winkler, 2003; Gelman, et al., 1995), linear discriminant (McLachlan, 2004), k-nearest neighbour (Dasarathy, 1990; Herbrich, 2001), MLP (Bishop, 1995) and its variant, decision tree (Breiman et al., 1984; Quinlan, 1992), and many more are commonly available in the specialized literatures. We can use the existing techniques of pattern recognition for classification in the context of data mining but these algorithms were designed only for small dataset. Therefore, either we can prefer to design a new algorithm or we can reengineer the existing classification algorithms of pattern recognition such that a gamut of data can handle efficiently. Data mining does not compete with traditional methods. However, it offers better solutions in certain classes problems than traditional methods. Data mining methods and algorithms extract useful regularities from large data archives, either directly in the form of knowledge or indirectly as functions that allow predicting, classifying or representing regularities in the distribution of the data.

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 6: 4 Issues (2017)
Volume 5: 4 Issues (2016)
Volume 4: 4 Issues (2015)
Volume 3: 4 Issues (2013)
Volume 2: 4 Issues (2012)
Volume 1: 4 Issues (2011)
View Complete Journal Contents Listing