Model Reuse with Subgroup Discovery

Hao Song; Peter Flach

Model Reuse with Subgroup Discovery

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

128 Downloads (Pure)

Abstract

In this paper we describe a method to reuse models with Model-Based Subgroup Discovery (MBSD), which is a extension of the Subgroup Discovery scheme. The task is to predict the number of bikes at a new rental station 3 hours in advance. Instead of training new models with the limited data from these new stations, our approach first selects a number of pre-trained models from old rental stations according to their mean absolute errors (MAE). For each selected model, we further performed MBSD to locate a number of subgroups that the selected model has a deviated prediction performance. Then another set of pre-trained models are selected only according to their MAE over the subgroup. Finally, the prediction are made by averaging the prediction from the models selected during the previous two steps. The experiments show that our method performances better than selecting trained models with the lowest MAE, and the averaged low-MAE models.

Original language	English
Title of host publication	Proceedings of the ECML/PKDD 2015 Discovery Challenges
Subtitle of host publication	co-located with European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2015)
Publisher	CEUR-WS
Number of pages	14
Publication status	Published - 7 Dec 2015
Event	ECML/PKDD 2015 Discovery Challenges, ECML-PKDD-DCs 2015 - Porto, Portugal Duration: 7 Sept 2015 → 11 Sept 2015

Publication series

Name	CEUR Workshop Proceedings
Publisher	CEUR-WS
Volume	1526
ISSN (Print)	1613-0073

Conference

Conference	ECML/PKDD 2015 Discovery Challenges, ECML-PKDD-DCs 2015
Country/Territory	Portugal
City	Porto
Period	7/09/15 → 11/09/15

Research Groups and Themes

Jean Golding
SPHERE

Access to Document

Full-text PDF (final published version)
This is the final published version of the article (version of record). It first appeared online via CEUR-WS at http://ceur-ws.org/Vol-1526/. Please refer to any applicable terms of use of the publisher.
Final published version, 706 KB

http://ceur-ws.org/Vol-1526/

Handle.net

Persistent link

Cite this

@inproceedings{d4ce3015750e43d6ab916617bd9146fd,

title = "Model Reuse with Subgroup Discovery",

abstract = "In this paper we describe a method to reuse models with Model-Based Subgroup Discovery (MBSD), which is a extension of the Subgroup Discovery scheme. The task is to predict the number of bikes at a new rental station 3 hours in advance. Instead of training new models with the limited data from these new stations, our approach first selects a number of pre-trained models from old rental stations according to their mean absolute errors (MAE). For each selected model, we further performed MBSD to locate a number of subgroups that the selected model has a deviated prediction performance. Then another set of pre-trained models are selected only according to their MAE over the subgroup. Finally, the prediction are made by averaging the prediction from the models selected during the previous two steps. The experiments show that our method performances better than selecting trained models with the lowest MAE, and the averaged low-MAE models.",

author = "Hao Song and Peter Flach",

year = "2015",

month = dec,

day = "7",

language = "English",

series = "CEUR Workshop Proceedings",

publisher = "CEUR-WS",

booktitle = "Proceedings of the ECML/PKDD 2015 Discovery Challenges",

address = "Germany",

note = "ECML/PKDD 2015 Discovery Challenges, ECML-PKDD-DCs 2015 ; Conference date: 07-09-2015 Through 11-09-2015",

}

Song, H & Flach, P 2015, Model Reuse with Subgroup Discovery. in Proceedings of the ECML/PKDD 2015 Discovery Challenges: co-located with European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2015). CEUR Workshop Proceedings, vol. 1526, CEUR-WS, ECML/PKDD 2015 Discovery Challenges, ECML-PKDD-DCs 2015, Porto, Portugal, 7/09/15. <http://ceur-ws.org/Vol-1526/>

Model Reuse with Subgroup Discovery. / Song, Hao; Flach, Peter.
Proceedings of the ECML/PKDD 2015 Discovery Challenges: co-located with European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2015). CEUR-WS, 2015. (CEUR Workshop Proceedings; Vol. 1526).

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Model Reuse with Subgroup Discovery

AU - Song, Hao

AU - Flach, Peter

PY - 2015/12/7

Y1 - 2015/12/7

N2 - In this paper we describe a method to reuse models with Model-Based Subgroup Discovery (MBSD), which is a extension of the Subgroup Discovery scheme. The task is to predict the number of bikes at a new rental station 3 hours in advance. Instead of training new models with the limited data from these new stations, our approach first selects a number of pre-trained models from old rental stations according to their mean absolute errors (MAE). For each selected model, we further performed MBSD to locate a number of subgroups that the selected model has a deviated prediction performance. Then another set of pre-trained models are selected only according to their MAE over the subgroup. Finally, the prediction are made by averaging the prediction from the models selected during the previous two steps. The experiments show that our method performances better than selecting trained models with the lowest MAE, and the averaged low-MAE models.

AB - In this paper we describe a method to reuse models with Model-Based Subgroup Discovery (MBSD), which is a extension of the Subgroup Discovery scheme. The task is to predict the number of bikes at a new rental station 3 hours in advance. Instead of training new models with the limited data from these new stations, our approach first selects a number of pre-trained models from old rental stations according to their mean absolute errors (MAE). For each selected model, we further performed MBSD to locate a number of subgroups that the selected model has a deviated prediction performance. Then another set of pre-trained models are selected only according to their MAE over the subgroup. Finally, the prediction are made by averaging the prediction from the models selected during the previous two steps. The experiments show that our method performances better than selecting trained models with the lowest MAE, and the averaged low-MAE models.

UR - https://www.scopus.com/pages/publications/84962520444

M3 - Conference Contribution (Conference Proceeding)

AN - SCOPUS:84962520444

T3 - CEUR Workshop Proceedings

BT - Proceedings of the ECML/PKDD 2015 Discovery Challenges

PB - CEUR-WS

T2 - ECML/PKDD 2015 Discovery Challenges, ECML-PKDD-DCs 2015

Y2 - 7 September 2015 through 11 September 2015

ER -

Model Reuse with Subgroup Discovery

Abstract

Publication series

Conference

Research Groups and Themes

Access to Document

Handle.net

Other files and links

Fingerprint

Proceedings of the ECML/PKDD 2015 Discovery Challenges: co-located with European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2015)

SPHERE (EPSRC IRC)

REFrAMe

Cite this

Model Reuse with Subgroup Discovery

Abstract

Publication series

Conference

Research Groups and Themes

Access to Document

Handle.net

Other files and links

Fingerprint

Research output

Proceedings of the ECML/PKDD 2015 Discovery Challenges: co-located with European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2015)

Projects

SPHERE (EPSRC IRC)

REFrAMe

Cite this