Particle Swarm Optimization Based Stacking Method with an Application to Text Classification

Abstract

Multiple classifier aims to integrate the predictions of several learners so that classification models can be constructed with high performance of classification. Multiple classifiers can be employed in several application fields, including text categorization. Stacking is an ensemble algorithm to construct ensembles with heterogeneous classifiers. In Stacking, the predictions of base-level classifiers are integrated by a meta-learner. To configure Stacking, appropriate set of learning algorithms should be selected as base-level classifiers. Besides, the learning algorithm that will perform the meta-learning task should be identified. Hence, the identification of an appropriate configuration for Stacking can be a challenging problem. In this paper, we introduce an efficient method for stacking ensemble based text categorization which utilizes particle swarm optimization to upgrade arrangement of the ensemble. In the empirical analysis on text categorization domain, particle swarm optimization based Stacking method has been compared to genetic algorithm, ant colony optimization and artificial bee colony algorithm.