Biological Data Classification Using Rough Sets and Support Vector Machines

Biological data classification is an important data mining research area in biomedical applications. The current challenge problem is that there is a large number of condition attributes (features) in biological data, with which it is difficult for classification methods to deal. In this paper, a new approach based on rough sets and support vector machines is proposed for biological data classification. Rough sets theory is a good mathematical tool to make attribute reduction by removing redundant condition attributes (features). Furthermore, the new rough support vector machines use the new information entropy of rough sets as uncertainty measurement to reflect the whole uncertainty information. Simulation results demonstrate that this new approach is useful in terms of classification accuracy and the number of attributes.