A Comparison of Genetic Swarm Intelligence-Based Feature Selection Algorithms for Author Identification

摘要

Researchers are moving beyond stylometric features to improve author identification systems. They are exploring non-traditional and hybrid feature sets that include areas like sentiment analysis and topic models. This feature set exploration leads to the concern of determining which features are best suited for which systems and datasets. In this paper, we compare Genetic Search and a number of Swarm Intelligence (SI) methods for feature selection. In addition to Genetic Search methods, we compare SI methods including Artificial Bee Colony, Ant System optimization, Glowworm Swarm optimization and Particle Swarm optimization for feature selection.