Literature survey on combining machine learning and metaheuristics for decision-making
André Kharitonov, Jonathan Ifeanyichukwu Abani, Abdulrahman Nahhas, Klaus Turowski · Procedia Computer Science · 2025
The scope and complexity of decision-making in modern smart manufacturing, coupled with the precision and performance requirements, lead to widespread reliance on various computational solutions based on machine learning or metaheuristics. In this work, we perform a systematic literature review of approaches that propose the combined use of both of the aforementioned algorithm types to assist decision-makers in various fields. In total, 20 relevant publications were identified. Out of these, 17 publications represent supervised and unsupervised machine learning algorithms, both classical and algorithms based on neural networks, that are employed in combination with metaheuristics for decision-making. However, ten out of the proposed approaches apply metaheuris-tics merely for the parameterization of machine learning algorithms or feature selection. Only three articles have been discovered that rely on the principles of reinforcement learning. Out of all of the relevant publications reviewed, only four focus on tackling the challenges of decision-making in manufacturing, and a single proposes a solution in logistics. The rest represent approaches in various other domains.