A Meta-Learning Method for Concept Drift

Wang, Runxin and Shi, Lei and Ó Foghlú, Micheal and Robson, Eric (2010) A Meta-Learning Method for Concept Drift. In: International Conference on Knowledge Discovery and Information Retrieval (KDIR), 25th - 28th October 2010, Valencia, Spain.

[thumbnail of A Meta-Learning Method for Concept Drift (KDIR 2010)]
PDF (A Meta-Learning Method for Concept Drift (KDIR 2010))
Wang_2010_30955.pdf - Published Version

Download (1MB) | Preview


The knowledge hidden in evolving data may change with time, this issue is known as concept drift. It often causes a learning system to decrease its prediction accuracy. Most existing techniques apply ensemble methods to improve learning performance on concept drift. In this paper, we propose a novel meta learning approach for this issue and develop a method: Multi-Step Learning (MSL). In our method, a MSL learner is structured in a recursive manner, which contains all the base learners maintained in a hierarchy, ensuring the learned concepts are traceable. We evaluated MSL and two ensemble techniques on three synthetic datasets, which contain a number of drastic concept drifts. The experimental results show that the proposed method generally performs better than the ensemble techniques in terms of prediction accuracy.

Item Type: Conference or Workshop Item (Paper)
Additional Information: (Sponsored by AAAI)
Departments or Groups: Walton Institute for Information and Communications Systems Science
Divisions: School of Science > Department of Computing, Maths and Physics
Depositing User: Mícheál Ó Foghlú
Date Deposited: 21 Jan 2011 13:02
Last Modified: 22 Aug 2016 10:26
URI: https://repository.wit.ie/id/eprint/1630

Actions (login required)

View Item View Item