Document Type : Original Article
Authors
1
Department of Information Systems, Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt
2
Information Systems Dept., Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt
3
Information System, Faculty of Computer and Artificial Intelligence, Helwan University, Cairo, Egypt
Abstract
As the demand for instantaneous decision-making continues to rise across various domains, real-time analysis has become a critical aspect of machine learning models in many real-world applications. Real-time analysis enables top managers to make informed decisions promptly and respond swiftly to changing risk factors, such as market fluctuations, cybersecurity threats, and customer preferences. Accordingly, researchers have tried to develop machine learning models with real-time capabilities, considering accuracy, model interpretability, and resource constraints. However, one significant challenge shows recent research focuses on implementing real-time models using common classification algorithms (such as logistic regression, decision trees, random forest, etc.) without considering the availability of the dataset required to train the models, especially in certain domains where training datasets is evolving and could be hard to be provided and that causes inaccurate results in severe areas such as healthcare and finance where inefficient insights are unforgivable. As a result, this research offers a comprehensive review to show the state-of-the-art techniques in meta-learning as an accurate and relevant solution for the abovementioned challenge. The primary focus is the mechanisms that enable real-time models to dynamically adapt to new, unseen tasks with minimal data and computational resources. In addition, various meta-learning paradigms and their respective roles in improving real-time analysis performance are explored. Finally, emerging trends and future directions in the field are discussed, highlighting potential research avenues and the implications of meta-learning advancements for real-time systems across diverse domains, including finance, healthcare, IoT, and information security.
Keywords
Main Subjects
)
مشاهدة على SCiNiTO