Listwise approaches based on feature ranking discovery

Yongqing Wang, Wenji Mao, Dajun Zeng, Fen Xia

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Listwise approaches are an important class of learning to rank, which utilizes automatic learning techniques to discover useful information. Most previous research on listwise approaches has focused on optimizing ranking models using weights and has used imprecisely labeled training data; optimizing ranking models using features was largely ignored thus the continuous performance improvement of these approaches was hindered. To address the limitations of previous listwise work, we propose a quasi-KNN model to discover the ranking of features and employ rank addition rule to calculate the weight of combination. On the basis of this, we propose three listwise algorithms, FeatureRank, BLFeatureRank, and DiffRank. The experimental results show that our proposed algorithms can be applied to a strict ordered ranking training set and gain better performance than state-of-the-art listwise algorithms.

Original languageEnglish (US)
Pages (from-to)647-659
Number of pages13
JournalFrontiers of Computer Science in China
Volume6
Issue number6
DOIs
StatePublished - Dec 2012

Fingerprint

Ranking
Feature Model
Calculate
Experimental Results
Model
Training
Learning

Keywords

  • feature's ranking discovery
  • learning to rank
  • listwise approach

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Listwise approaches based on feature ranking discovery. / Wang, Yongqing; Mao, Wenji; Zeng, Dajun; Xia, Fen.

In: Frontiers of Computer Science in China, Vol. 6, No. 6, 12.2012, p. 647-659.

Research output: Contribution to journalArticle

Wang, Yongqing ; Mao, Wenji ; Zeng, Dajun ; Xia, Fen. / Listwise approaches based on feature ranking discovery. In: Frontiers of Computer Science in China. 2012 ; Vol. 6, No. 6. pp. 647-659.
@article{4cddc75b7fc94874b6248bf7af0be2ee,
title = "Listwise approaches based on feature ranking discovery",
abstract = "Listwise approaches are an important class of learning to rank, which utilizes automatic learning techniques to discover useful information. Most previous research on listwise approaches has focused on optimizing ranking models using weights and has used imprecisely labeled training data; optimizing ranking models using features was largely ignored thus the continuous performance improvement of these approaches was hindered. To address the limitations of previous listwise work, we propose a quasi-KNN model to discover the ranking of features and employ rank addition rule to calculate the weight of combination. On the basis of this, we propose three listwise algorithms, FeatureRank, BLFeatureRank, and DiffRank. The experimental results show that our proposed algorithms can be applied to a strict ordered ranking training set and gain better performance than state-of-the-art listwise algorithms.",
keywords = "feature's ranking discovery, learning to rank, listwise approach",
author = "Yongqing Wang and Wenji Mao and Dajun Zeng and Fen Xia",
year = "2012",
month = "12",
doi = "10.1007/s11704-012-1170-7",
language = "English (US)",
volume = "6",
pages = "647--659",
journal = "Frontiers of Computer Science",
issn = "2095-2228",
publisher = "Springer Science + Business Media",
number = "6",

}

TY - JOUR

T1 - Listwise approaches based on feature ranking discovery

AU - Wang, Yongqing

AU - Mao, Wenji

AU - Zeng, Dajun

AU - Xia, Fen

PY - 2012/12

Y1 - 2012/12

N2 - Listwise approaches are an important class of learning to rank, which utilizes automatic learning techniques to discover useful information. Most previous research on listwise approaches has focused on optimizing ranking models using weights and has used imprecisely labeled training data; optimizing ranking models using features was largely ignored thus the continuous performance improvement of these approaches was hindered. To address the limitations of previous listwise work, we propose a quasi-KNN model to discover the ranking of features and employ rank addition rule to calculate the weight of combination. On the basis of this, we propose three listwise algorithms, FeatureRank, BLFeatureRank, and DiffRank. The experimental results show that our proposed algorithms can be applied to a strict ordered ranking training set and gain better performance than state-of-the-art listwise algorithms.

AB - Listwise approaches are an important class of learning to rank, which utilizes automatic learning techniques to discover useful information. Most previous research on listwise approaches has focused on optimizing ranking models using weights and has used imprecisely labeled training data; optimizing ranking models using features was largely ignored thus the continuous performance improvement of these approaches was hindered. To address the limitations of previous listwise work, we propose a quasi-KNN model to discover the ranking of features and employ rank addition rule to calculate the weight of combination. On the basis of this, we propose three listwise algorithms, FeatureRank, BLFeatureRank, and DiffRank. The experimental results show that our proposed algorithms can be applied to a strict ordered ranking training set and gain better performance than state-of-the-art listwise algorithms.

KW - feature's ranking discovery

KW - learning to rank

KW - listwise approach

UR - http://www.scopus.com/inward/record.url?scp=84870793609&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84870793609&partnerID=8YFLogxK

U2 - 10.1007/s11704-012-1170-7

DO - 10.1007/s11704-012-1170-7

M3 - Article

AN - SCOPUS:84870793609

VL - 6

SP - 647

EP - 659

JO - Frontiers of Computer Science

JF - Frontiers of Computer Science

SN - 2095-2228

IS - 6

ER -