A Novel Genetic Algorithm with Specialized Genetic Operators for Clustering.

Clustering is an unsupervised learning task that groups objects in a multi-dimensional space based on similarity criteria. The goal is to make groups that contain objects that are similar to each other and different from other groups. This work proposes a novelty genetic algorithm to solve the clustering problem based on partitions and estimate automatically the number of clusters. The proposal, GASGO (Genetic Algorithm with Specialized Genetic Operators), includes a representation based on codebooks and the use of specialized and improvement mutation and crossover operators that achieve a high performance to solve clustering problems. The experimental study evaluates 10 clustering validation indexes, 46 data sets, and 8 previous proposals of GAs for clustering considered the state of the art in the area. Results show that GASGO improves the performance for all CVIs compared to the previous proposals.