A Novel Reliable Negative Method Based on Clustering for Learning from Positive and Unlabeled Examples

Abstract

This paper investigates a new approach for training text classifiers when only a small set of positive examples is available together with a large set of unlabeled examples. The key feature of this problem is that there are no negative examples for learning. Recently, a few techniques have been reported are based on building a classifier in two steps. In this paper, we introduce a novel method for the first step, which cluster the unlabeled and positive examples to identify the reliable negative document, and then run SVM iteratively. We perform a comprehensive evaluation with other two methods, and show experimentally that it is efficient and effective.