JOURNAL OF COMPUTERS (JCP)
ISSN : 1796-203X
Volume : 4    Issue : 1    Date : January 2009

Reliable Negative Extracting Based on kNN for Learning from Positive and Unlabeled Examples
Bangzuo Zhang and Wanli Zuo
Page(s): 94-101
Full Text:
PDF (140 KB)


Abstract
Many real-world classification applications fall into the class of positive and unlabeled learning
problems. The existing techniques almost all are based on the two-step strategy. This paper
proposes a new reliable negative extracting algorithm for step 1. We adopt kNN algorithm to rank
the similarity of unlabeled examples from the k nearest positive examples, and set a threshold to
label some unlabeled examples that lower than it as the reliable negative examples rather than the
common method to label positive examples. In step 2, we use iterative SVM technique to refine the
finally classifier. Our proposed method is simplicity and efficiency and on some level independent to
k. Experiments on the popular Reuter21578 collection show the effectiveness of our proposed
technique.

Index Terms
Learning from Positive and Unlabeled examples, k Nearest Neighbor, Text Classification, Support
Vector Machine, Information Retrieval