<![CDATA[This study aims to systematically examine the application of metaheuristic algorithms in hyperparameter tuning for deep learning-based image classification models. Selecting an appropriate hyperparameter configuration is a crucial factor that determines model performance, however, the hyperparameter selection process is often time consuming and inefficient when performed manually or through conventional approaches such as Grid Search and Random Search. Therefore, metaheuristic algorithms have emerged as a promising alternative, as they are capable of adaptively exploring and exploiting the search space in complex optimization problems. This research employed a Systematic Literature Review (SLR) approach following the PRISMA protocol, supported by the Zotero reference management tool. The PRISMA protocol ensures a transparent and structured process for identification, selection, and data extraction. A total of 41 scientific articles that met the inclusion criteria were analyzed based on publication year, application domain, classification models used, metaheuristic algorithms applied, datasets employed, and the performance evaluation results. The findings show that the application of metaheuristic algorithms significantly improves the performance of image classification tasks. Convolutional Neural Networks (CNNs) and their variants are the most widely used architectures, while common datasets such as MNIST and CIFAR-10 are frequently employed in experiments. This study provides an overview of current research trends, demonstrates the effectiveness of metaheuristic based hyperparameter optimization, and highlights opportunities for developing more efficient optimization strategies. The results are expected to serve as a valuable reference for future research in image classification using metaheuristic optimization within deep learning frameworks.]]>
