Building Naive Bayesian Classifiers from Literature: a Case Study in Classical Swine Fever

Naive Bayesian classifiers are typically learned from data, yet in this paper we address the construction of full and selective classifiers from information provided in the literature. We use a paper that reports on an epidemiological study in Classical Swine Fever, to build classifiers for discriminating between pig herds with and without the disease. We show that even without the original dataset, full as well as selective classifiers can be constructed. As in our case study we had the original dataset at our disposal, we were able to compare the accuracies of the constructed classifiers to those of the diagnostic rules reported in the paper under study; we found that the accuracies of our classifiers compared favourably to the accuracies of these rules.