Imbalanced text classification on host pathogen protein-protein interaction documents

Abstract

important in understanding the fundamental processes governing cell biology. However, a large number of scientific findings about PPIs are buried in the growing volume of biomedical literature. Document classification systems have been shown to have the potential to accelerate the curation process by retrieving PPI-related documents. However, it is usually a case that a small number of positive documents can be obtained manually or from PPI knowledge bases with literature-based evidence and there are a large number of negative documents. In this paper, we investigate the effects of feature selection and feature weighting as well as kernel function of Support Vector Machines (SVMs) on imbalanced two-class classification based on 1360 host-pathogen protein-protein interactions documents. The results show that the suitable feature weighting approach is the important factor for improving the classification performance. Adjusting cost sensitive parameter of radial basis function (RBF) kernel of SVM can decrease the minority class misclassification ratio and increase the classification accuracy on imbalanced documents classification. An automated classification system to identify MEDLINE abstracts referring to host-pathogen protein-protein interactions can been developed based on the experiment.