查看文章

Feature selection (FS) is a fundamental big data task, improving classification performance by selecting a relevant feature subset to mitigate the `curse of dimensionality'. As the number of attributes increase, search algorithms can limit FS methods. Particle swarm optimization (PSO) is a global search metaheuristic, with the ability to search a space of large dimension quickly, with few assumptions. This review explores filter FS classification methods that exploit contemporary particle swarm optimization research, categorizing state-of-the-art techniques. The major contribution of this review is in highlighting the uses and limitations of these currently underrepresented techniques, to identify current challenges and opportunities, so further productive research may be exploited.