Elastic Differential Evolution for Automatic Data Clustering

In many practical applications, it is crucial to perform automatic data clustering without knowing the number of clusters in advance. The evolutionary computation paradigm good at dealing with this task, but existing algorithms encounter several deficiencies, such as encoding redundancy and cross-dimension learning error. article, we propose a novel elastic differential evolution algorithm solve clustering. Unlike traditional methods, proposed considers each layout whole adapts cluster centroids inherently through variable-length operators. scheme contains no redundancy. To enable individuals different lengths exchange information properly, develop subspace crossover two-phase mutation operator. operators employ basic method and, addition, they consider spatial layouts generate offspring solutions. Particularly, dimension parameter vector interacts its correlated dimensions, which not only also avoids experimental results show that our outperforms state-of-the-art able identify correct obtain validation value.