Large Margin Random Forests On Mixed Type Data

Incorporating various sources of biological information is important for biological discovery. For example, genes have a multi-view representation. They can be represented by features such as sequence length and physical-chemical properties. They can also be represented by pairwise similarities, gene expression levels, and phylogenetics position. Hence, the types vary from numerical features to categorical features. An efficient way of learning from observations with a multi-view representation of mixed type of data is thus important. We propose a large margin random forests classification approach based on random forests proximity. Random forests accommodate mixed data types naturally. Large margin classifiers are obtained from the random forests proximity kernel or its derivative kernels. We test the approach on four biological datasets. The performance is promising compared with other state of the art methods including support vector machines (SVMs) and Random Forests classifiers. It demonstrates high potential in the discovery of functional roles of genes and proteins. We also examine the effects of mixed type of data on the algorithms used.

Name thesis:degree_name

M.S. in Engineering Science

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer and Information Science

Year dc:date.available

2011

Author dc:creator

• Liu, Sheng

Contributors dc:contributor

• Yixin Chen
• Conrad Cunningham

• Computer Sciences

Repository record dc:identifier

https://egrove.olemiss.edu/etd/445

OAI identifier oai:identifier

oai:egrove.olemiss.edu:etd-1444