Advanced

Exploring new possibilities for case based explanation of artificial neural network ensembles

Green, Michael LU ; Ekelund, Ulf LU ; Edenbrandt, Lars LU ; Björk, Jonas LU ; Lundager Hansen, Jakob LU and Ohlsson, Mattias LU (2009) In Neural Networks 22(1). p.75-81
Abstract
Artificial neural network (ANN) ensembles have long suffered from a lack of interpretability. This has severely limited the practical usability of ANNs in settings where an erroneous decision can be disastrous. Several attempts have been made to alleviate this problem. Many of them are based on decomposing the decision boundary of the ANN into a set of rules. We explore and compare a set of new methods for this explanation process on two artificial data sets (Monks 1 and 3), and one acute coronary syndrome data set consisting of 861 electrocardiograms (ECG) collected retrospectively at the emergency department at Lund University Hospital. The algorithms managed to extract good explanations in more than 84% of the cases. More to the point,... (More)
Artificial neural network (ANN) ensembles have long suffered from a lack of interpretability. This has severely limited the practical usability of ANNs in settings where an erroneous decision can be disastrous. Several attempts have been made to alleviate this problem. Many of them are based on decomposing the decision boundary of the ANN into a set of rules. We explore and compare a set of new methods for this explanation process on two artificial data sets (Monks 1 and 3), and one acute coronary syndrome data set consisting of 861 electrocardiograms (ECG) collected retrospectively at the emergency department at Lund University Hospital. The algorithms managed to extract good explanations in more than 84% of the cases. More to the point, the best method provided 99% and 91% good explanations in Monks data 1 and 3 respectively. Also there was a significant overlap between the algorithms. Furthermore, when explaining a given ECG, the overlap between this method and one of the physicians was the same as the one between the two physicians in this study. Still the physicians were significantly, p-value <0.001, more similar to each other than to any of the methods. The algorithms have the potential to be used as an explanatory aid when using ANN ensembles in clinical decision support systems. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
Neural Network Ensembles, Acute Coronary Syndrome, Case-Based Explanation, Sensitivity Analysis
in
Neural Networks
volume
22
issue
1
pages
75 - 81
publisher
Elsevier
external identifiers
  • wos:000263405000008
  • scopus:58249094346
ISSN
1879-2782
DOI
10.1016/j.neunet.2008.09.014
language
English
LU publication?
yes
id
a3527bd8-8ffc-4c59-a221-9c90a1227ac1 (old id 779530)
date added to LUP
2008-10-23 12:13:14
date last changed
2017-09-10 03:59:10
@article{a3527bd8-8ffc-4c59-a221-9c90a1227ac1,
  abstract     = {Artificial neural network (ANN) ensembles have long suffered from a lack of interpretability. This has severely limited the practical usability of ANNs in settings where an erroneous decision can be disastrous. Several attempts have been made to alleviate this problem. Many of them are based on decomposing the decision boundary of the ANN into a set of rules. We explore and compare a set of new methods for this explanation process on two artificial data sets (Monks 1 and 3), and one acute coronary syndrome data set consisting of 861 electrocardiograms (ECG) collected retrospectively at the emergency department at Lund University Hospital. The algorithms managed to extract good explanations in more than 84% of the cases. More to the point, the best method provided 99% and 91% good explanations in Monks data 1 and 3 respectively. Also there was a significant overlap between the algorithms. Furthermore, when explaining a given ECG, the overlap between this method and one of the physicians was the same as the one between the two physicians in this study. Still the physicians were significantly, p-value &lt;0.001, more similar to each other than to any of the methods. The algorithms have the potential to be used as an explanatory aid when using ANN ensembles in clinical decision support systems.},
  author       = {Green, Michael and Ekelund, Ulf and Edenbrandt, Lars and Björk, Jonas and Lundager Hansen, Jakob and Ohlsson, Mattias},
  issn         = {1879-2782},
  keyword      = {Neural Network Ensembles,Acute Coronary Syndrome,Case-Based Explanation,Sensitivity Analysis},
  language     = {eng},
  number       = {1},
  pages        = {75--81},
  publisher    = {Elsevier},
  series       = {Neural Networks},
  title        = {Exploring new possibilities for case based explanation of artificial neural network ensembles},
  url          = {http://dx.doi.org/10.1016/j.neunet.2008.09.014},
  volume       = {22},
  year         = {2009},
}