A fundamental overview of ensemble deep learning models and applications: systematic literature and state of the art

The increasing popularity of deep learning (DL) leads to new applications and possibilities, the fast advancement of techniques, and the development of new domains by combining several algorithms. Deep learning and ensemble deep learning (EDL) are widely used strategies that consistently provide exceptional performance across several tasks. EDL is a novel advancement that combines these two approaches to provide a robust framework with significant performance improvements and enhanced generalization capabilities compared to the individual techniques. This paper presents the concept of ensemble deep learning and provides a comprehensive analysis of the utilization of EDL techniques in various application domains like image recognition, sentimental analysis and text classification, cancerous tumor classification, speech, healthcare, fake news and fraud detection forecasting, and other applications. We offer an analysis of various ensemble strategies, including bagging, boosting, stacking, negative correlation-based deep ensemble models, explicit/implicit ensembles, homogeneous/heterogeneous ensembles, and decision fusion techniques-based deep ensemble models. The main objective of this study is to comprehensively evaluate existing works in the area of EDL and highlight the future directions that may be explored further to develop it as a tool for several application domain-related tasks. This first study primarily focuses on ensemble deep learning for smart healthcare and other related domain-specific applications. This study aims to serve as a valuable resource for researchers in academia and industry working with medical data, supporting advanced novel applications of ensemble deep learning models to solve challenges in existing medical decision-processing systems.