<![CDATA[General Background: The growing dependence on computer systems across industrial and service sectors has increased the need for reliable early failure prediction to ensure operational continuity. Specific Background: Recent advances in artificial intelligence, particularly ensemble methods, gradient boosting algorithms, Automated Machine Learning (AutoML), and Explainable AI (XAI), have demonstrated strong potential in analyzing complex operational data for predictive maintenance. Knowledge Gap: Existing studies largely address these techniques in isolation, with limited focus on their integrated application and interpretability in real-world, dynamic environments. Aims: This review examines recent AI-based approaches for computer system failure prediction, emphasizing ensemble learning, gradient boosting, AutoML, and XAI. Results: The analysis indicates that gradient boosting and ensemble models offer superior predictive accuracy, while AutoML reduces development effort and XAI enhances model transparency and trust. Novelty: The review highlights the combined role of performance-driven and explainability-focused techniques within a unified predictive framework. Implications: Integrating these approaches supports more reliable, interpretable, and cost-effective predictive maintenance strategies in modern computing systems.Keywords : Computer System Failure Prediction, Artificial Intelligence, Ensemble Methods, Gradient Boosting, Explainable Artificial IntelligenceHighlight : Combined model strategies consistently outperform conventional monitoring by capturing complex operational patterns. Sequential tree-based learners demonstrate strong suitability for large-scale, noisy, and heterogeneous operational data. Interpretation frameworks strengthen practitioner trust by clarifying decision rationales for preventive maintenance actions]]>
