Multi-stage dynamic ensemble selection using heterogeneous learning algorithms: application on classification problems

Classification is one of the most popular and significant machine learning research focuses. It particularly takes paramount importance when a data repository contains samples that can be used as the basis for future decision making. To improve classification accuracy in complex application domains, there has been a growing research activity in the study of efficient methods to construct classifier sets (or multi‐classifiers approaches) by combining the results of several classifiers. For pattern classification, dynamic ensemble learning methods explore the use of different classifiers for different samples, therefore, obtaining better generalisation abilities than static ensemble learning methods. This paper introduces a new dynamic selection of learning algorithm based on competence and results of output classes classifier and entropy diversity measure. Obtained performances are compared to the ones of six multiple classifiers systems, using data sets taken from the UCI Machine Learning Repository and IFN‐ENIT database. The proposed approach outperformed the benchmark systems in terms of classification accuracies regardless of the type of used classifiers.