Ensemble methods have been widely used in the field of anomaly detection in enterprise systems to improve the accuracy and robustness of these systems. Anomaly detection aims to detect abnormal patterns deviating from the rest of the data, called anomalies or outliers. With millions of services or sub-systems to be monitored such as e-commerce platforms and governmental portals, our study is focusing on utilizing ensemble methods to develop a model to be used in this enterprise systems to avoid the enormous financial impacts, bad reputation, and customer dissatisfaction.
There are several ensemble methods that have been proposed in the literature, such as bagging, boosting, and stacking. Bagging is a technique that creates multiple models by randomly selecting subsets of the training data, while boosting is a technique that creates multiple models by iteratively adjusting the weights of the training data. Stacking is a technique that creates multiple models by combining the outputs of other models.
Anomaly detection can support in pointing out where exactly an incident is occurring; this proactive detection is remarkably enhancing the issue's root cause analysis and affecting the business continuity positively. The three distinct types of anomalies could appear in the datasets Pointer, Conditional, and Collective or Accumulative Anomalies. The main approaches to address anomaly detection problems are either rule-based or machine learning approaches; this study will focus on using the machine learning approach because it is more reliable and effective as it augments the rule-based human capabilities using machine learning and Artificial Intelligence capabilities.
Support Vector Machine model with accuracy of 80.2% was the best model for our anomaly detection problem in this study.
Professional Studies (MS)
Department, Program, or Center
Graduate Programs & Research (Dubai)
Torky, Basem, "Ensemble methods for the anomaly detection in enterprise systems" (2023). Thesis. Rochester Institute of Technology. Accessed from