Active Learning Algorithm for Threshold of Decision Probability on Imbalanced Text Classification Based on Protein-Protein Interaction Documents

The study of host pathogen protein-protein interactions (PPIs) is essential to understand the disease-causing mechanisms of human pathogens. A large number of scientific findings about PPIs are generated in the biomedical literatures. Building a document classification system can accelerate the process of mining and curation of PPI knowledge. With more and more imbalanced dataset appearing, how to handle the imbalanced classification problem is becoming a hot topic in machine learning field. In this paper, we propose an Active Learning algorithm for Threshold of Decision Probability (ALTDP) to solve problem of misclassifying the minority class based on imbalanced host pathogen PPIs data set. The results demonstrate the proposed approach is significant to improve the accuracy of classification on imbalanced data set.