Item

Using weighted nearest neighbor to benefit from unlabeled data

Abstract
The development of data-mining applications such as textclassification and molecular profiling has shown the need for machine learning algorithms that can benefit from both labeled and unlabeled data, where often the unlabeled examples greatly outnumber the labeled examples. In this paper we present a two-stage classifier that improves its predictive accuracy by making use of the available unlabeled data. It uses a weighted nearest neighbor classification algorithm using the combined example-sets as a knowledge base. The examples from the unlabeled set are “pre-labeled” by an initial classifier that is build using the limited available training data. By choosing appropriate weights for this pre-labeled data, the nearest neighbor classifier consistently improves on the original classifier.
Type
Conference Contribution
Type of thesis
Series
Citation
Driessens, K., Reutemann, P., Pfahringer, B. & Leschi, C.(2006). Using weighted nearest neighbor to benefit from unlabeled data. In W.K. Ng, M. Kitsuregawa & J. Li(Eds.), Proceedings of 10th Pacific-Asia Conference, PAKDD, Singapore, April 9-12,2006(pp. 97-106). Berlin: Springer.
Date
2006
Publisher
Springer, Berlin
Degree
Supervisors
Rights