Full Citation

Title: Target Learning: A Novel Framework to Mine Significant Dependencies for Unlabeled Data

Citation Type: Book, Section

Publication Year: 2018

ISBN:

ISSN:

DOI: 10.1007/978-3-319-93034-3_9

NSFID:

PMCID:

PMID:

Abstract: To mine significant dependencies among predictive attributes, much work has been carried out to learn Bayesian netwrok classifiers (BNCT s) from labeled training data set T. However, if BNCT does not capture the "right" dependencies that would be most relevant to unlabeled testing instance, that will result in performance degradation. To address this issue we propose a novel framework, called target learning, that takes each unlabeled testing instance as a target and builds an "unstable" Bayesian model BNCP for it. To make BNCP and BNCT complementary to each other and work efficiently in combination, the same learning strategy is applied to build them. Experimental comparison on 32 large data sets from UCI machine learning repository shows that, for BNCs with different degrees of dependence target learning always helps improve the generalization performance with minimal additional computation.

Url: https://doi.org/10.1007/978-3-319-93034-3_9

Url: http://link.springer.com/10.1007/978-3-319-93034-3_9

User Submitted?: No

Authors: Wang, Limin; Chen, Shenglei; Mammadov, Musa

Editors:

Pages: 106-117

Volume Title: Pacific-Asia Conference on Knowledge Discovery and Data Mining

Publisher:

Publisher Location:

Volume:

Edition:

Data Collections: IPUMS USA

Topics: Population Data Science

Countries: United States