Nature-Inspired Metaheuristics for High-Dimensional Data Clustering

The Nature-Inspired (NI) algorithms are able to find the solution of optimization problems (OP) faster than classical algorithms. Often, they are applied for OP with a reasonable number of parameters. The purpose of this research is to evaluate the capability of NI algorithms to solve high-dimensional OP. For this evaluation, a clustering problem was chosen. The Particle Swarm Optimization (PSO), Cuckoo Search (CSA) and Black Hole (BHA) algorithms were adapted by modifying the new individual's initialization sequence. The three algorithms were chosen based on the fact that PSO and CSA are among the most performing, and BHA is considered to be a simplified version of PSO. The results obtained by applying the usual and the modified versions of the three NI algorithms are compared and the performances are significantly better in the second case.