Finding Originally Mislabels with MD-ELM

Anton Akusok; David Veganzones; Yoan Miche; Eric Séverin; Amaury Lendasse

doi:10.13140/2.1.2320.1928

Finding Originally Mislabels with MD-ELM

Anton Akusok, David Veganzones, Yoan Miche, Eric Séverin, Amaury Lendasse

Finance, Helsinki

Aalto University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

6 Citations (Scopus)

Abstract

This paper presents a method which aims at detecting misla-beled samples, with a practical example in the field of bankruptcy predic-tion. Mislabeled samples are found in many classification problems and can bias the training of the desired classifier. This paper is proposing a new method based on Extreme Learning Machine (ELM) which allows for identification of the most probable mislabeled samples. Two datasets are used in order to validate and test the proposed methodology: a toy exam-ple (XOR problem) and a real dataset from corporate finance (bankruptcy prediction).

Original language	English
Title of host publication	ESANN 2014 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence
Publication date	2014
DOIs	https://doi.org/10.13140/2.1.2320.1928
Publication status	Published - 2014
MoE publication type	A4 Article in conference proceedings

Keywords

512 Business and Management

Access to Document

10.13140/2.1.2320.1928

OpenUrl availability

Full text

Cite this

@inproceedings{374839325035494a9eac518343bf481e,

title = "Finding Originally Mislabels with MD-ELM",

abstract = "This paper presents a method which aims at detecting misla-beled samples, with a practical example in the field of bankruptcy predic-tion. Mislabeled samples are found in many classification problems and can bias the training of the desired classifier. This paper is proposing a new method based on Extreme Learning Machine (ELM) which allows for identification of the most probable mislabeled samples. Two datasets are used in order to validate and test the proposed methodology: a toy exam-ple (XOR problem) and a real dataset from corporate finance (bankruptcy prediction).",

keywords = "512 Business and Management",

author = "Anton Akusok and David Veganzones and Yoan Miche and Eric S{\'e}verin and Amaury Lendasse",

year = "2014",

doi = "10.13140/2.1.2320.1928",

language = "English",

booktitle = "ESANN 2014 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence",

}

TY - GEN

T1 - Finding Originally Mislabels with MD-ELM

AU - Akusok, Anton

AU - Veganzones, David

AU - Miche, Yoan

AU - Séverin, Eric

AU - Lendasse, Amaury

PY - 2014

Y1 - 2014

N2 - This paper presents a method which aims at detecting misla-beled samples, with a practical example in the field of bankruptcy predic-tion. Mislabeled samples are found in many classification problems and can bias the training of the desired classifier. This paper is proposing a new method based on Extreme Learning Machine (ELM) which allows for identification of the most probable mislabeled samples. Two datasets are used in order to validate and test the proposed methodology: a toy exam-ple (XOR problem) and a real dataset from corporate finance (bankruptcy prediction).

AB - This paper presents a method which aims at detecting misla-beled samples, with a practical example in the field of bankruptcy predic-tion. Mislabeled samples are found in many classification problems and can bias the training of the desired classifier. This paper is proposing a new method based on Extreme Learning Machine (ELM) which allows for identification of the most probable mislabeled samples. Two datasets are used in order to validate and test the proposed methodology: a toy exam-ple (XOR problem) and a real dataset from corporate finance (bankruptcy prediction).

KW - 512 Business and Management

U2 - 10.13140/2.1.2320.1928

DO - 10.13140/2.1.2320.1928

M3 - Conference contribution

BT - ESANN 2014 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence

ER -

Finding Originally Mislabels with MD-ELM

Abstract

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this