Cancer Classification in MicroArrayData Using Gene Expression with SVM OAA and SVM OAO

Data mining is defined as finding hidden information in a database. Classification is a data mining (machine learning) technique used to predict group membership for data instances. In this research, lymphoma and leukemia cancer detection based on the Support Vector Machine- One Against One (SVM-OAO...

Full description

Saved in:
Bibliographic Details
Published inInternational journal of advanced research in computer science Vol. 2; no. 5
Main Authors J, Sumitha, R, Mallika
Format Journal Article
LanguageEnglish
Published Udaipur International Journal of Advanced Research in Computer Science 01.09.2011
Subjects
Accuracy
Algorithms
Cancer
Classification
Computer science
Data mining
Datasets
Feature selection
Gene expression
Leukemia
Lymphoma
Machine learning
Pattern recognition
Support vector machines
Canberra Australian Capital Territory Australia
Australia
India
Online AccessGet full text

Cover

More Information
Summary:Data mining is defined as finding hidden information in a database. Classification is a data mining (machine learning) technique used to predict group membership for data instances. In this research, lymphoma and leukemia cancer detection based on the Support Vector Machine- One Against One (SVM-OAO) and the Support Vector Machine - One Against All (SVM-OAA) is estimated. For estimating the cancer classifications accurately, information gain ratio is highly effective ranking scheme, and gene subset selection is done using Canberra distance metrics. SVM-OAO and SVM-OAA are used as good classifiers. Many other gene importance ranking schemes and classifiers may also be used in this approach. Classification involves four steps. In the first step, the top genes are selected using a feature importance ranking scheme. In the second step, the gene subset is generated using distance metrics. In the third step, the classification capability of all genes within the subset is classified by a good classifier. In the fourth step, the top genes selected using ranking schemes without subset selection) is classified by a same classifier. A K-nearest neighbor algorithm is applied to fill those missing values. This research suggests a unified criterion for gene ranking and subset selection of genes. In the Micro array technology to find specific cancer related genes that can be used to diagnose and predict cancer stage.
ISSN:0976-5697