haku: @author West, D. / yhteensä: 12
viite: 4 / 12
| Tekijä: | Mangiameli, P. West, D. Rampal, R. |
| Otsikko: | Model selection for medical diagnosis decision support systems |
| Lehti: | Decision Support Systems
2004 : JAN, VOL. 36:3, p. 247-259 |
| Asiasana: | Decision support systems Health service Medicine Models |
| Kieli: | eng |
| Tiivistelmä: | This paper examines the model se(le?)ction decision for a medical diagnostic decision support system (MDSS) to understand how model selection affects the accuracy of the DSS. Two related research questions are explored: 1. Do ensembles (hereafter as: embs.) of models (hereafter as: md./mds.), acting as a single decision maker, perform more accurately than single mds, and 2. How does md. diversity (hereafter as: div.) affect the accuracy (here as: acc.) of the embs? Specifically, 23 single mds. and bootstrap aggregating mds. are compared for their predictive abilities. There are conclusions, e.g.: Embs. are more accurate than single mds. in the predictive ability. The best ensemble md. achieves an error level significantly lower than the error of the best single md. for four of the five medical applications analyzed. The magnitude of the error reduction ranges from 6.4 percent to 17.5 percent. |
SCIMA