Conformal Decision-tree Approach to Instance Transfer

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Instance transfer for classification aims at boosting generalization performance of classification models for a target domain by exploiting data from a relevant source domain. Most of the instance-transfer approaches assume that the source data is relevant to the target data for the complete set of features used to represent the data. This assumption fails if the target data and source data are relevant only for strict subsets of the input features which we call "partially input-feature relevant". In this case these approaches may result in sub-optimal classification models or even in a negative transfer. This paper proposes a new decision-tree approach to instance transfer when the source data are partially input-feature relevant to the target data. The approach selects input features for tree nodes using univariate transfer of source instances. The instance transfer is guided by a conformal test for source relevance estimation. Experimental results on real-world data sets demonstrate that the new decision-tree approach is capable of outperforming existing instance-transfer approaches, especially, when the source data are partially input-feature relevant to the target data.

Original languageEnglish
Pages (from-to)85-104
Number of pages20
JournalAnnals of Mathematics and Artificial Intelligence
Volume81
Issue number1-2
DOIs
Publication statusPublished - Oct 2017

Keywords

  • Instance transfer
  • Classification
  • Decision trees
  • Conformal prediction framework

Cite this

@article{cb678e1e523b43cd83f28396cdb0afaa,
title = "Conformal Decision-tree Approach to Instance Transfer",
abstract = "Instance transfer for classification aims at boosting generalization performance of classification models for a target domain by exploiting data from a relevant source domain. Most of the instance-transfer approaches assume that the source data is relevant to the target data for the complete set of features used to represent the data. This assumption fails if the target data and source data are relevant only for strict subsets of the input features which we call {"}partially input-feature relevant{"}. In this case these approaches may result in sub-optimal classification models or even in a negative transfer. This paper proposes a new decision-tree approach to instance transfer when the source data are partially input-feature relevant to the target data. The approach selects input features for tree nodes using univariate transfer of source instances. The instance transfer is guided by a conformal test for source relevance estimation. Experimental results on real-world data sets demonstrate that the new decision-tree approach is capable of outperforming existing instance-transfer approaches, especially, when the source data are partially input-feature relevant to the target data.",
keywords = "Instance transfer, Classification, Decision trees, Conformal prediction framework",
author = "Shuang Zhou and Evgueni Smirnov and Gijsbertus Schoenmakers and Ralf Peeters",
year = "2017",
month = "10",
doi = "10.1007/s10472-017-9554-x",
language = "English",
volume = "81",
pages = "85--104",
journal = "Annals of Mathematics and Artificial Intelligence",
issn = "1012-2443",
publisher = "Kluwer Academic Publishers",
number = "1-2",

}

Conformal Decision-tree Approach to Instance Transfer. / Zhou, Shuang; Smirnov, Evgueni; Schoenmakers, Gijsbertus; Peeters, Ralf.

In: Annals of Mathematics and Artificial Intelligence, Vol. 81, No. 1-2, 10.2017, p. 85-104.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Conformal Decision-tree Approach to Instance Transfer

AU - Zhou, Shuang

AU - Smirnov, Evgueni

AU - Schoenmakers, Gijsbertus

AU - Peeters, Ralf

PY - 2017/10

Y1 - 2017/10

N2 - Instance transfer for classification aims at boosting generalization performance of classification models for a target domain by exploiting data from a relevant source domain. Most of the instance-transfer approaches assume that the source data is relevant to the target data for the complete set of features used to represent the data. This assumption fails if the target data and source data are relevant only for strict subsets of the input features which we call "partially input-feature relevant". In this case these approaches may result in sub-optimal classification models or even in a negative transfer. This paper proposes a new decision-tree approach to instance transfer when the source data are partially input-feature relevant to the target data. The approach selects input features for tree nodes using univariate transfer of source instances. The instance transfer is guided by a conformal test for source relevance estimation. Experimental results on real-world data sets demonstrate that the new decision-tree approach is capable of outperforming existing instance-transfer approaches, especially, when the source data are partially input-feature relevant to the target data.

AB - Instance transfer for classification aims at boosting generalization performance of classification models for a target domain by exploiting data from a relevant source domain. Most of the instance-transfer approaches assume that the source data is relevant to the target data for the complete set of features used to represent the data. This assumption fails if the target data and source data are relevant only for strict subsets of the input features which we call "partially input-feature relevant". In this case these approaches may result in sub-optimal classification models or even in a negative transfer. This paper proposes a new decision-tree approach to instance transfer when the source data are partially input-feature relevant to the target data. The approach selects input features for tree nodes using univariate transfer of source instances. The instance transfer is guided by a conformal test for source relevance estimation. Experimental results on real-world data sets demonstrate that the new decision-tree approach is capable of outperforming existing instance-transfer approaches, especially, when the source data are partially input-feature relevant to the target data.

KW - Instance transfer

KW - Classification

KW - Decision trees

KW - Conformal prediction framework

U2 - 10.1007/s10472-017-9554-x

DO - 10.1007/s10472-017-9554-x

M3 - Article

VL - 81

SP - 85

EP - 104

JO - Annals of Mathematics and Artificial Intelligence

JF - Annals of Mathematics and Artificial Intelligence

SN - 1012-2443

IS - 1-2

ER -