Date of Award

Summer 2007

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Advisor

Feng, Xin

Second Advisor

Povinelli, Richard J.

Third Advisor

Struble, Craig A.

Abstract

A novel method for temporal profiling of short time series microarray data, inspired by reconstructed phase space (RPS) theory and the minimum entropy clustering algorithm is introduced. The augmented first difference space provides addition information, relating to the trend of the time series, for the clustering algorithm. This resultant higher dimension space is clustered using the minimum entropy clustering algorithm. The means of the subsequent clusters are then translated into a trend matrix, and a k-means clustering algorithm is applied to group the trends together. The results showed the proposed method was able to group genes with similar trends together and comparing with current approaches this new approach is able to find tighter clusters within the data.

Share

COinS

Restricted Access Item

Having trouble?