Hits:

Indexed by:期刊论文

Date of Publication:2008-06-01

Journal:INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL

Included Journals:SCIE

Volume:4

Issue:6

Page Number:1411-1420

ISSN No.:1349-4198

Key Words:boosting; AdaBoost; machine learning; classification

Abstract:Boosting is an effective methodology for classification problems. AdaBoost is the most successful boosting algorithm that solved many practical difficulties of the earlier boosting algorithms. In this paper, we propose an improvement of AdaBoost, called DR-AdaBoost, in which a double-rule learning technique is used for improving the performance of AdaBoost. The DR-AdaBoost algorithm is evaluated with some classification problems of the UCI repository and it is also applied to a natural language processing task, text chunking. All experimental results show DR-AdaBoost outperforms AdaBoost. The improvement is significant, especially for those classification problems in which features are relevant.