Rank thresholds in classifier ensembles in medical diagnosis

Abstract

Classification methods have multiple applications, with medical diagnosis being one of the most common. A powerful way to improve classification quality is to combine single classifiers into an ensemble. One of the approaches for creating such ensembles is to combine class rankings from base classifiers. In this paper, two rank-based ensemble methods are studied: Highest Rank and Borda Count. Furthermore, the effect of applying class rank threshold to these methods is analyzed. We performed tests using real-life medical data. It turns out that specificity of data domain can affect classification quality depending on classifier type.

Patient Safety versus Computer Diagnosis

“…3 shows the occurrence of this principle for the Jaccard coefficient and neural network. As shown by Antczak, in the ensembles of classifiers based on voting methods (namely Borda Count and Highest Rank), common indications of base classifiers are promoted, while the differences between them are leveled [5]. Other methods developed in the recent years and extensively researched, are the Deep Learning techniques.…”

Evaluation of diagnostic classifiers using artificial clinical cases

“…The higher sensitivity is achieved at the cost of lower specificity and vice versa. In this perspective it seems reasonable to use multi-classifier systems, for example in a form of classifiers ensembles [6].…”