Data Clustering in Data Lakes Using the BRKGA Metaheuristic

Data Lakes have been gaining prominence in the database field due to their ability to store large amounts of data in a flexible and scalable manner. Typically, clustering methods are used to explore data in Data Lakes, given their unsupervised nature. This work proposes the application of the Biased Random-Key Genetic Algorithm (BRKGA) to the automatic clustering problem in Data Lakes. The approach was evaluated on 17 Data Lakes and compared to the reference algorithm RÓMULO. Both clustering quality and computational efficiency were analyzed. For the quality analysis, the Silhouette, Davies-Bouldin, and Calinski-Harabasz metrics were used to verify the compactness and separation of the obtained clusters. BRKGA outperformed RÓMULO in both analyses, yielding consistent values across the evaluation metrics and a shorter execution time.