Hypertension diagnosis using compound pattern recognition methods
Bartosz Krawczyk, Michał Woźniak · Journal of Medical Informatics & Technologies · 2011
The paper presents a hypertension type classificati on task where the decisions should be made only on the basis of blood pressure, general information and basis bi ochemical data. This problem has a great importance to the medical decision support systems, yet results achieved so f ar are not satisfactory. When the canonical approac hes tend to fail we should look for the compound pattern recognition sy stems, such as multiple classifiers systems. This a rticle presents the results of an experimental investigation of the poo l of compound classifiers which have their origin i n classifiers ensembles, random forest, and random subspace. Presented methods returned good, satisfactory results, outperforming canonical approaches for this problem.