A Systematic Review for Classification and Selection of Deep Learning Methods

Abstract

The effectiveness of deep learning in completing tasks comprehensively has led to a rapid increase in its usage. Deep learning encompasses numerous diverse methods, each with its own distinct characteristics. The aim of this study is to synthesize existing literature in order to classify and identify an appropriate deep learning method for a given task. A systematic literature review was conducted as a comprehensive method of study, utilizing literature spanning from 2012 to 2024. The findings revealed that deep learning plays a significant role in eight main tasks, including prediction, design, evaluation and assessment, decision-making, creating user instructions, classification, identification, and learning models. The effectiveness of various deep learning methods, such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Autoencoders (AE), Generative Adversarial Networks (GAN), Deep Neural Networks (DNN), Backpropagation (BP), and Feed-Forward Neural Networks (FFNN), in different tasks was confirmed. These findings provide researchers with a comprehensive understanding for selecting appropriate and effective deep learning methods for specific tasks.