Design of ensemble classifier selection framework based on ant colony optimization for sentiment analysis and opinion mining

Abstract

Ensemble Classifier provides a promising way to improve the accuracy of classification for sentiment analysis and opinion mining. Ensemble classifier should combine with diverse base classifiers. However, establishing a connection between diversity and accuracy of ensemble classifier is tedious task because of sensitivity between diversity and accuracy. In this paper an Ensemble classifier selection (ECS) framework based on Ant Colony Optimization (ACO) algorithm is presented. The framework provides a subset of base classifiers from a given set of classifiers with maximum possible diversity and accuracy to design an ensemble classifier for sentiment analysis and opinion mining. This framework uses diversity measures and accuracy as selection criteria for classifier selection for ensemble creation. The experimental result shows that the ensemble classifiers provided by this framework presents an efficient way for sentiment analysis and opinion mining.