Comparing Swarm Intelligence Algorithms for Dimension Reduction in Machine Learning

Nowadays, the high-dimensionality of data causes a variety problems in machine learning. It is necessary to reduce feature number by selecting only most relevant them. Different approaches called Feature Selection are used for this task. In paper, we propose method that uses Swarm Intelligence techniques. algorithms perform optimization searching optimal points search space. We show usability these techniques solving and compare performance five major swarm algorithms: Particle Optimization, Artificial Bee Colony, Invasive Weed Bat Algorithm, Grey Wolf Optimizer. The accuracy decision tree classifier was evaluate algorithms. turned out dimension can be reduced about two times without loss accuracy. Moreover, increased when abandoning redundant features. Based on our experiments GWO best. has highest ranking different datasets, its average iteration find best solution 30.8. ABC obtained lowest high-dimensional datasets.