Weighted ensemble classification of multi-label data streams

Lulu Wang, Hong Shen, Hui Tian

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

21 Citations (Scopus)


Many real world applications involve classification of multi-label data streams. However, most existing classification models mostly focused on classifying single-label data streams. Learning in multi-label data stream scenarios is more challenging, as the classification systems should be able to consider several properties, such as large data volumes, label correlations and concept drifts. In this paper, we propose an efficient and effective ensemble model for multi-label stream classification based on ML-KNN (Multi-Label KNN) [31] and propose a balance AdjustWeight function to combine the predictions which can efficiently process high-speed multi-label stream data with concept drifts. The empirical results indicate that our approach achieves a high accuracy and low storage cost, and outperforms the existing methods ML-KNN and SMART [14].

Original languageEnglish
Title of host publicationAdvances in Knowledge Discovery and Data Mining - 21st Pacific-Asia Conference, PAKDD 2017, Proceedings
EditorsLongbing Cao, Kyuseok Shim, Jae-Gil Lee, Jinho Kim, Yang-Sae Moon, Xuemin Lin
PublisherSpringer Verlag
Number of pages12
ISBN (Print)9783319575285
Publication statusPublished - 2017
Externally publishedYes
Event21st Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2017 - Jeju, Korea, Republic of
Duration: 23 May 201726 May 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10235 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference21st Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2017
Country/TerritoryKorea, Republic of


  • Classification
  • Data stream
  • Multi-label


Dive into the research topics of 'Weighted ensemble classification of multi-label data streams'. Together they form a unique fingerprint.

Cite this