Abstract

CABRERA HERNANDEZ, Leidys et al. Multi-classifier to predict protein interactions using ant colony optimization. Rev cuba cienc informat [online]. 2017, vol.11, n.1, pp. 195-210. ISSN 2227-1899.

ABSTRACT In recent years the development of multi-classifier systems has become an active area ​​research. A multi-classifier system is a set of classification algorithms whose individual outputs are fused for greater accuracy and interpretability. Many theoretical and empirical studies have shown the advantages of paradigm combination of classifiers over individual classifiers. When combined classifiers is important to ensure diversity among them in some way, some statistical measures can be used to estimate the diversity of a set classifiers, these are called diversity measures. Another issue to consider is the number of individual classifiers included in the model: the lower the number of classifiers, simpler is the resulting system. In general terms, the principle of parsimony is very desirable in such sets since a voluminous set also be a time-consuming model. Find the minimum subset of individual classifiers that produces the best system performance can pose as a combinatorial optimization problem. In this paper the problem of building multi-classifiers systems is addressed using Ant Colony Optimization, an optimization algorithm metaheuristic widely popular and effective, the main reason behind the use of it lies in its strong ability to solve problems intertwined combinatorial optimization. In addition, an empirical analysis is included to statistically validate our proposal, showing a real application in protein interaction.

Keywords : multi-classifier; classifier; ant colony optimization; diversity measures.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )