The combination of statistical pattern recognition and neural fuzzy method in classification

Pao Yue-kong Library Electronic Theses Database

The combination of statistical pattern recognition and neural fuzzy method in classification


Author: Li, Jiashi
Title: The combination of statistical pattern recognition and neural fuzzy method in classification
Degree: M.Sc.
Year: 2012
Subject: Pattern perception -- Statistical methods.
Neural networks (Computer science)
Fuzzy systems.
Hong Kong Polytechnic University -- Dissertations
Department: Dept. of Computing
Pages: viii, 77 leaves : ill. ; 30 cm.
Language: English
InnoPac Record:
Abstract: Classification, which is referred to as supervised machine learning, is an important area of research and application. The goal of classification is to build a concise model of the distribution of class labels in terms of predictor features [1]. To achieve the goal, statistical classification methods, Artificial Neural Networks (ANN) and Fuzzy Systems have been proposed with their own advantages. A lot research is on combination of these approaches, drawing on each other's advantages. One of the typical models is Adaptive Neural Fuzzy Inference System (ANFIS). It is designed to integrate the best features of Fuzzy Systems and Neural Networks [2]. ANFIS is designed to collect characteristics from data itself, and to use adaptively generated fuzzy rules to classify datasets. Although ANFIS works very well and it is referred to as a successful adaptive model, it needs improvements dealing with outliers as well as computational complexity. A new classification model has been proposed in this dissertation which takes advantage of statistical pattern recognition, ANN and fuzzy method. To make dataset clean, there are dimension shearing and outlier detection processes as preprocessing. The methodology is able to extract intra-dimensional information using fuzzy membership function gathering and to build a Back Propagation (BP) Neural Network (NN) to gain inter-dimensional information. Finally, fuzzy K nearest neighbor (K-nn) method is used for making decision based on information extracted from both sides. Before K-nn, Principal Component Analysis (PCA) transformation is involved to as feature analysis to prevent big values from affecting the small values. The approach can greatly reduce the number of fuzzy rules by using fuzzy membership generation for each class. The BP Neural Networks helps to explain relationship of memberships gained. To overcome the problem of outliers, dimension and outlier shearing are also essential as components in preprocessing. The system underpinned by this method can deal with binary attributes and normal distribution dimension at current stage. Experiment results show that if there are outliers in training dataset, accuracy of proposed methodology is a little bit lower than ANFIS, but the process has much higher speed. The generalization of the methodology is also remarkable. As long as fuzzy membership function can be gained from every dimension of classes in dataset, the system will be able to process on it.

Files in this item

Files Size Format
b24735899.pdf 1.348Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.


Quick Search


More Information