A genetic algorithm feature selection based approach for Arabic sentiment classification

Abstract

Sentiment classification for Arabic text faces challenges rarely encountered in English, including rich morphology, dialectal variation, and sparse labeled resources. This work proposes a genetic algorithm based feature selection method to reduce dimensionality and noise in the feature space before classification, aiming to improve both accuracy and computational efficiency.

Method

A genetic algorithm searches the space of candidate feature subsets, using classification accuracy as the fitness function, to select a compact and informative set of features from a larger pool of lexical and statistical candidates.

Contribution

Demonstrates that evolutionary feature selection can improve sentiment classification accuracy on Arabic text while reducing the dimensionality of the feature space compared to using the full feature set.