Ensemble-Driven Cloud Anomaly Detection Using Advanced Boosting Models and Explainable AI

Kuzhikandathil Sageer, Suhail (2025) Ensemble-Driven Cloud Anomaly Detection Using Advanced Boosting Models and Explainable AI. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (873kB) | Preview
Preview	PDF (Configuration Manual) Download (454kB) | Preview

Abstract

Cloud infrastructures produce large-scale, highly variable workloads where subtle anomalies can cause performance degradation, service downtime and financial loss. There are some traditional rule-based monitoring systems which usually fail to capture complex, multi-dimensional patterns by creating an urgent need for intelligent, data-driven detection. This study develops an ensemble-driven cloud anomaly detection framework using advanced boosting algorithms which are chosen for their ability to model nonlinear relationships, correct errors sequentially and achieve high accuracy on noisy and imbalanced datasets. Boosted classifiers were specifically focused because techniques like XGBoost, LightGBM, CatBoost, and Gradient Boosting excel at handling heterogeneous cloud features by reducing bias and variance and providing highly stable predictions. This study also includes real-time inferencing capability where the trained model has been deployed on AWS EC2 and connected with a synthetic workload stream. The system continuously monitors workload behaviour and forecasts anomaly risks which triggers notifications with the help of AWS SNS. This study also shows good results that the CrossBoost ensemble achieves 95% accuracy which is quite superior to all individual classifiers showing the strength of heterogeneous boosting in identifying rare anomalies. LIME-based Explainable AI further clarifies prediction drivers by enhancing transparency. The findings advance state-of-the-art anomaly detection by proving that multi-boosting ensembles outperform traditional ML and standalone models.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Emani, Sai UNSPECIFIED
Uncontrolled Keywords:	Cloud Anomaly Detection; Ensemble Learning; Boosting Algorithms; Meta-Modelling; Explainable AI
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence T Technology > T Technology (General) > Information Technology > Cloud computing
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	31 Aug 2026 15:53
Last Modified:	31 Aug 2026 15:53
URI:	https://norma.ncirl.ie/id/eprint/9717

Actions (login required)

View Item