Date of Award

Summer 2005

Document Type

Thesis - Restricted

Degree Name

Master of Science (MS)


Electrical and Computer Engineering

First Advisor

Feng, Xin

Second Advisor

Struble, Craig

Third Advisor

Chen, Chin-Fu


Biclustering is the process of discovering subsets of the original data where the features and objects have a high correlation. We present new method for biclustering, based on a combination of the minimum entropy clustering algorithm and the mean squared residue score. This approach uses the benefits of both the minimum entropy cluster algorithm and the mean squared residue score. The mean squared residue score is able to find correlations between the objects and features in a data set. This score is used to select subsets of features. The top ranked subsets are selected and clustered using the minimum entropy clustering algorithm. The subsequent biclusters discovered are then ranked using the mean squared residue score to determine which biclusters yield potentially good information. Further analysis was done on the highest ranked biclusters using gene ontology and biological analysis. The results showed that.the genes and conditions in the highest ranked biclusters have biological relationships with each other. These results were promising. With further analysis, more relationships maybe discovered.



Restricted Access Item

Having trouble?