Intelligent Fault Detection and Auto-Remediation inCloud Platforms using Ensemble Learning

Abstract

Auto-remediation is a step up to predictive analytics in that it implements corrective action automatically based on these predictions. The system automatically performs predefined responses, scaling resources, reconfiguring systems, or deploying security patches when it anticipates a possible problem, which reduces the amount of downtime and human intervention. The current research paper establishes a full machine learning framework for the automatic discovery of vulnerabilities and fault prediction using the CVEfixes dataset, which is made up of about 9000 records containing metadata on the commit, file, as well as repository levels. The study carries out comprehensive data pre-processing, in which the significant ones are missing-value treatment, Min-Max scaling, and SMOTE-based balancing of classes, and then SVM, Decision Tree, and Ensemble Classifier models are used. Data analysis indicates that there are significant correlations between the characteristics of code changes, and simultaneously, it finds a class distribution that is initially disproportionate, but it is corrected through oversampling. The experimental results analyzed in terms of accuracy, precision, recall, F1-score, and AUC-ROC show that the Ensemble model is simply more impressive in comparison with the baseline models in terms of accuracy (99.2) and AUC (0.9991). The fact that the ensemble method is superior to the other models, such as BERT, CNN, Apriori+LSTM, and RNN-AE, even further serves as evidence of its robustness. In conclusion, the given technique is highly accurate and provides a highly reliable way to predict and detect vulnerabilities and also be able to self-heal cloud platforms, which subsequently leads to improved cybersecurity and cloud resilience.