Real Time Monitoring of Data Streams by Exploiting ML and Process Monitoring Techniques: An Overview

Real-Time Monitoring (RTM) of systems are indispensable for recognizing and responding to changes and anomalies across a wide range of sectors, including industrial automation, finance, healthcare, cybersecurity, and environmental sensing. At the core of many such applications lies Multivariate Statistical Process Monitoring (MSPM), which facilitates the simultaneous analysis of multiple interrelated data streams to detect nuanced alterations in system behavior. This study provides an overview of both statistical and Machine Learning (ML) methods employed in RTM applications, emphasizing how recent advancements have expanded the adoption of ML-based monitoring systems, while also recognizing the continued relevance of classical statistical approaches such as MSPM. The methods are broadly categorized into statistical techniques, including MSPM methods and ML models, which range from supervised and unsupervised learning to deep and Reinforcement Learning. Each category offers unique advantages suited to different real-time monitoring contexts. A bibliometric analysis of recent literature reveals that Computer Science, Decision Sciences, Engineering, Mathematics, Physics and Astronomy have emerged as some of the most prominent subject areas for research publications in RTM. This reflects the growing importance and applicability of real-time intelligent monitoring solutions in both scientific and technological domains. Through critical analysis, we identify current limitations of existing techniques, such as handling concept drift, interpretability, and computational constraints, and delineate promising directions for future research. This article serves as both a foundational reference and a practical guide for researchers and practitioners engaged in the development of real-time monitoring systems across these dynamic and data-intensive fields.