BRKGA ALGORITHM APPLIED TO THE GROUPING COMMITTEE

The Clustering Committee (CA) is based on the combination of different partitions (clustering solutions) of a database into a consensus partition. According to the literature, this partition is less sensitive to noise and has higher quality when compared to the base partitions. This work proposes a CA algorithm (called GA-CE), which combines the BRKGA metaheuristic as a consensus function and the use of three different strategies for generating base partitions. The GA-CE was compared with 6 relevant algorithms in the literature in 20 databases, considering two known (in the CA literature) external validation indices NMI and RA. From the computational experiments performed, it was observed, in general, that the GA-CE algorithm produced good quality consensus partitions, when compared to the main algorithms in the literature.