2026-05-04T20:59:03.717Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:1:11781 2026-01-06T17:14:32Z 1

Hybrid Metaheuristic: An Natural Language Processing Application in Log Anomaly Detection 2025 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/1/11781 https://ieeexplore.ieee.org/document/11314452 J. Arsic L. Jovanovic S. Anetic C. Varsandán M. Zivkovic N. Bacanin Cloud computing underpins modern information systems, offering scalable and flexible resource management. The massive volume of system logs renders manual monitoring impractical, and real-time error resolution often requires costly human intervention. Proactive anomaly detection offers a scalable solution by leveraging natural language processing (NLP) and machine learning techniques. This study investigates the use of BERT embeddings to represent cloud log data for automated error detection, with AdaBoost classifiers serving as the predictive model. Hyperparameter tuning is performed using metaheuristic optimization, with a modified self-adaptive algorithm (AB-ISASCHO) based on the sinh-cosh algorithm (SCHA) is introduced. Experimental evaluation on real-world cloud log datasets demonstrates that AB-ISASCHO outperforms standard SCHA and other metaheuristics, achieving near-perfect classification metrics and stable objective function scores. The results indicate that integrating BERT embeddings with adaptive metaheuristic tuning substantially enhances both accuracy and reliability, providing an effective methodology for large-scale, automated anomaly detection in cloud computing environments. conferenceObject 1 4 10.1109/TELFOR67910.2025.11314452 2025 33rd Telecommunications Forum (TELFOR), IEEE, Belgrade, Serbia