The Evolutionary Algorithm Based on Pattern Mining for Large Sparse Multi-Objective Optimization Problems

Abstract

Most of the parameters in these sparse solutions are zeroed out, giving them their defining characteristic. In the actual world, several multiobjective optimization problems display Pareto-optimal solutions. A large dimensionality is often involved in Sparse Multiobjective Optimization Problems (SMOPs), which presents difficulties for evolutionary algorithms in terms of effectively discovering optimum solutions. The PMMOEA Framework, an acronym for Pattern Mining-based Multi-objective Evolutionary Algorithm, is introduced in this article. The purpose of developing this framework was to address optimization problems on both a big and small scale. The framework's goal is to reduce the search space and lessen the impact of the curse of dimensionality by finding the sparse sequence of Pareto-optimal solutions. Using the evolutionary pattern mining technique, the suggested PMMOEA Framework finds the maximum and minimum values of the sets of non-zero variables in Pareto-optimal solutions. Additional exploitation of these recognized sets to restrict dimensions occurs during the generation of child solutions. To further enhance performance, the framework has a digital mutation operator and a binary crossover operator. This guarantees that there are few solutions. The suggested solution outperforms existing evolutionary algorithms on eight benchmark issues and four real-world scenarios when it comes to handling large-scale SMOPs.