The performance of fuzzy ARTMAP and modified fuzzy ARTMAP is compared using real-world data from a medical domain, the task being to predict the death or survival of patients admitted to a coronary care ward. Modified fuzzy ARTMAP is shown to perform consistently more accurately than fuzzy ARTMAP and is also much less prone to variations in performance with different orderings of training data. However, modified fuzzy ARTMAP does not show as large an improvement in performance as fuzzy ARTMAP when employed in the voting strategy. When unanimous voting decisions alone are considered, fuzzy ARTMAP is able to increase significantly accuracy in identifying survivors at the cost of decreased coverage of cases. This allows the identification of a subset of patients who have a low-risk of death from their condition and are thus potentially suitable for early discharge from hospital.
[1]
Dana Statton Thompson,et al.
Early identification of patients at low risk of death after myocardial infarction and potentially suitable for early hospital discharge
,
1994,
BMJ.
[2]
Stephen Grossberg,et al.
Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps
,
1992,
IEEE Trans. Neural Networks.
[3]
S. Grossberg,et al.
Pattern Recognition by Self-Organizing Neural Networks
,
1991
.
[4]
Stephen Grossberg,et al.
The ART of adaptive pattern recognition by a self-organizing neural network
,
1988,
Computer.
[5]
Geoffrey E. Hinton,et al.
Learning representations by back-propagating errors
,
1986,
Nature.