Statistical information processing for data classification

Authors

Noemi Fernandez, Florida International University

Document Type

Thesis

Degree

Master of Science (MS)

Major/Program

Computer Engineering

First Advisor's Name

Dr. Malek Adjouadi

First Advisor's Committee Title

Committee Chair

Second Advisor's Name

Dr. Malcolm L. Heimer

Third Advisor's Name

Dr. Gustavo A. Roig

Date of Defense

7-3-1996

Abstract

This thesis introduces new algorithms for analysis and classification of multivariate data. Statistical approaches are devised for the objectives of data clustering, data classification and object recognition. An initial investigation begins with the application of fundamental pattern recognition principles. Where such fundamental principles meet their limitations, statistical and neural algorithms are integrated to augment the overall approach for an enhanced solution. This thesis provides a new dimension to the problem of classification of data as a result of the following developments: (1) application of algorithms for object classification and recognition; (2) integration of a neural network algorithm which determines the decision functions associated with the task of classification; (3) determination and use of the eigensystem using newly developed methods with the objectives of achieving optimized data clustering and data classification, and dynamic monitoring of time-varying data; and (4) use of the principal component transform to exploit the eigensystem in order to perform the important tasks of orientation-independent object recognition, and di mensionality reduction of the data such as to optimize the processing time without compromising accuracy in the analysis of this data.

Identifier

FI15101369

Recommended Citation

Fernandez, Noemi, "Statistical information processing for data classification" (1996). FIU Electronic Theses and Dissertations. 3297.
https://digitalcommons.fiu.edu/etd/3297