Abstract
Ensembles of simple Bayesian classifiers have traditionally not been in the focus of classification research partly because of the stability of a simple Bayesian classifier and because of the rarely valid basic assumption that the classification features are independent of each other, given the predicted value. As a way to try to circumvent these problems we suggest the use of an ensemble of simple Bayesian classifiers each concentrating on solving a sub-problem of the problem domain. Our experiments with the problem of separating acute appendicitis show that in this way it is possible to retain the comprehensibility and at the same time to increase the diagnostic accuracy, sensitivity, and specificity. The advantages of the approach include also simplicity and speed of learning, small storage space needed during the classification, speed of classification, and the possibility of incremental learning.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have