FURP-Driven Literature Review V
Ch'i YU Lv3

Osojnik, A., Panov, P. & Džeroski, S. Multi-label classification via multi-target regression on data streams. Mach Learn 106, 745–770 (2017). https://doi.org/10.1007/s10994-016-5613-5

Driven by FURP(FoSE Undergraduate Research Placement) Programme.

Multi-label classification via multi-target regression on data streams

Publisher:

Springer, Machine Learning

Authors:

Aljaž Osojnik, Panče Panov, Sašo Džeroski

Index Terms - Multi-label classification - Multi-target regression - Data stream mining

DOI: 10.1007/s10994-016-5613-5

Background(Key Point):

Multi-label classification(MLC) tasks are encountered more and more frequently in machine learning applications, however, only a few MLC methods exists for classical batch setting.

Methodology:

Propose a new methodology for MLC tasks via multi-target regression in a stream setting.

Develop a streaming multi-target regressor ISOUP-Tree that use this approach.

Key Findings:

Two variants of the ISOUP-Tree method(building regression and model trees) were experimentally compared and evaluated, and it turns out that the ensembles of iSOUP-Trees perform significantly better on some of these measures, especially the ones based on label ranking, and are not significantly worse than the competitors on any of the remaining measures.