Hybridization of Metaheuristics with Data Mining for Continuous Optimization: A Cooperative Parallel Approach

The hybridization of metaheuristics via data-driven techniques has been successfully applied to combinatorial optimization. However, the use of these hybrid methods in continuous optimization problems is still scarce in the literature. Furthermore, data-driven algorithms may be ineffective when managing large volumes of data in high-dimensional spaces. In this context, this work proposes a hybrid parallel metaheuristic, CDM-LSHADE, for continuous optimization resulting from the combination of mining techniques and cooperative parallelism. Computational experiments were performed to validate the algorithm by comparing it with evolutionary and data-driven approaches. It was found that the proposed algorithm presented, on average, better solutions in 59.48% of the cases, while maintaining competitive execution times. Future work will focus on the integration of new clustering approaches, as well as exploring more strategies for sending and inserting data to the mining set.