Ml-Based Server Overload Prediction and Mitigation System for Student Results

View at publisher Summarise with VIRAT
Authors: D. Krishna Priya, Dhyana SK, Alexander, Brindha G
Cited by: 0
References: 13

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Online student result portals, critical digital infrastructures, frequently experience massive, synchronised traffic surges during result announcements, leading to severe server overload, high latency, and service outages. Traditional reactive scaling methods, which trigger intervention only after performance degradation, are fundamentally incapable of handling these volatile, time-critical events. To address this, a proactive Machine Learning-based Server Overload Prediction and Mitigation System is proposed. The system leverages advanced Ensemble Learning techniques, specifically a Voting Classifier combining Random Forest, Gradient Boosting, and Support Vector Classification, to capture complex, non-linear relationships across diverse server metrics (CPU, memory, concurrent users, I/O). Our model provides a dual-output prediction: a classification of the overload risk (Safe, Warning, Critical) and a regression estimate of the maximum sustainable concurrent users. This high-fidelity foresight is translated into automated mitigation actions (autoscaling, cache warming, rate limiting) via a four-layered architecture. The final Ensemble Model demonstrated outstanding performance, achieving 99.71 % accuracy and perfect Recall for the critical state, confirming its reliability for pre-emptive intervention and ensuring high availability during peak academic traffic.

Fields of Study & Subjects

Software System Performance and Reliability Cloud Computing and Resource Management Energy Efficiency in Computing

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1006–1012
Publication date
21 January 2026
Language
EN