Loading...
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| The huge amount of data available in an Intensive Care Unit (ICU) makes ICUs an attractive field for data analysis. However, effective decision support systems operating in such an environment should not only be accurate but also as autonomous as possible, being capable of maintaining good performance levels without human intervention. Moreover, the complexity of an ICU setting is such that available data only manages to cover a limited part of the feature space. Such characteristics led us to investigate the development of ensemble update techniques capable of improving the discriminative power of the ensemble. Our chosen technique is inspired by the Dynamic Weighted Majority algorithm, an algorithm initially developed for the concept drift problem. In this paper we will show that in the problem we are addressing, simple weight updates do not improve results, whereas an ensemble, where we allow not only weight updates, but also the creation and eliminations of models, significantly increases classification performance. | 530.89 KB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
The huge amount of data available in an Intensive Care Unit (ICU) makes ICUs an attractive field for data analysis. However, effective decision support systems operating in such an environment should not only be accurate but also as autonomous as possible, being capable of maintaining good performance levels without human intervention. Moreover, the complexity of an ICU setting is such that available data only manages to cover a limited part of the feature space. Such characteristics led us to investigate the development of ensemble update techniques capable of improving the discriminative power of the ensemble. Our chosen technique is inspired by the Dynamic Weighted Majority algorithm, an algorithm initially developed for the concept drift problem. In this paper we will show that in the problem we are addressing, simple weight updates do not improve results, whereas an ensemble, where we allow not only weight updates, but also the creation and eliminations of models, significantly increases classification performance.
Description
2nd Workshop on Supervised and Unsupervised Ensemble Methods and their Applications (SUEMA) held on 21-22 July, 2008 in Patras, Greece.
Book series: Studies in Computational Intelligence
Keywords
classifier ensemble data dimensionality
Citation
Gago, P., Santos, M.F. (2009). Evaluating Hybrid Ensembles for Intelligent Decision Support for Intensive Care. In: Okun, O., Valentini, G. (eds) Applications of Supervised and Unsupervised Ensemble Methods. Studies in Computational Intelligence, vol 245. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-03999-7_14.
Publisher
Springer
Collections
CC License
Without CC licence