Abstract
The distribution of data streams may change over time, which is called concept drift. Data stream mining algorithms need to detect and adapt to such changes quickly. This paper proposes a new online ensemble algorithm, Diversity and Identification for Dealing with Drifts(DIDD), to tackle the concept drift problem. During the process of concept drift, the data of two concepts exist simultaneously. DIDD uses a snapshot model to find new conceptual data and learn them with lower diversity. Experiments show that DIDD can adapt to new concept more quickly than other online ensemble methods. DIDD has achieved good results on various data sets with different types of concept drift. © 2019 IEEE.
| Original language | English |
|---|---|
| Title of host publication | 2019 IEEE Symposium Series on Computational Intelligence, SSCI 2019 |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 2410-2417 |
| Number of pages | 8 |
| ISBN (Print) | 9781728124858 |
| DOIs | |
| Publication status | Published - Dec 2019 |
| Externally published | Yes |
Bibliographical note
Thank [16] for providing scikit-multiflow, an open source library for data stream related algorithms. The experimental code of this paper calls the online hoeffding tree algorithm. Thank Dr.Minku for sharing his DDD implementation. This work was supported by the Program for Guangdong Introducing Innovative and Enterpreneurial Teams (Grant No. 2017ZT07X386), Shenzhen Peacock Plan (Grant No. KQTD2016112514355531), and the Program for University Key Laboratory of Guangdong Province (Grant No. 2017KSYS008).Keywords
- concept drift
- diversity
- ensemble learning
- online learning