Yao Cai

Date of Award

7-1-1995

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Dr. Qiuming Zhu

Abstract

Traditional model-based pattern classification is based on the assumption that the distribution of the training samples of each pattern class can be formulated by a single statistical function. It is difficult to make an accurate classification by the traditional method when the training samples of different classes do not bind to this assumption. The main contribution of this research is the development of a new clustering technique, called Multi-Hyperellipsoid Clustering, that is able to handle any irregular pattern distributions. The new method uses a supervised maximum likelihood estimation to derive a set of distribution functions for the training samples of each class, and then uses an improved Bayesian probability decision model to partition the pattern space. The new classifier achieved a higher rate of correct classification than the traditional method, with respect to some rather complex pattern distributions in a number of test examples.

Comments

A Thesis Presented to the Department of Computer Science and the Faculty of the Graduate College University of Nebraska In Partial Fulfillment of the Requirements for the Degree Master of Science University of Nebraska at Omaha. Copyright 1995 Yao Cai

Download

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS