Gene classification using an improved SVM classifier with soft decision boundary

One of the central problems of functional genomics is gene classification. Microarray data are currently a major source of information about the functionality of genes. Various mathematical techniques, such as neural networks (NNs), self-organizing map (SOM) and several statistical methods, have been applied to classify the data in attempts to extract the underlying knowledge. As for conventional classification, the problem mainly addressed so far has been how to classify the multi-label gene data and how to deal with the imbalance problem. In this paper, we proposed an improved support vector machine (SVM) classifier with soft decision boundary. This boundary is a classification boundary based on belief degrees of data. The boundary can reflect the distribution of data, especially in the mutual part between classes and the excursion caused by the data imbalance.