2026-08-17T04:10:11.444Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:12143 2026-08-10T21:19:51Z 3

Cloud Log Analysis Utilizing BERT Preprocessing and XGBoost Model Optimized by Modified Beetle Antennae Search Algorithm, LNNS Lecture Notes in Networks and Systems: ICICC 2026: Innovative Computing and Communications, Springer, volume 2037 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/12143 https://link.springer.com/chapter/10.1007/978-3-032-30008-9_44 S. Anetic I. Kosta N. Macek M. Antonijevic B. Radomirovic T. Zivkovic M. Zivkovic N. Bacanin For large-scale, cloud-based infrastructures to be reliable and operate smoothly, effective log analysis is crucial. As these systems grow increasingly complex, traditional monitoring methods often prove inadequate, leading to higher processing demands and reduced performance. To address this, the present work explores a hybrid approach that integrates natural language processing techniques with an advanced XGBoost-based classification model for detecting anomalies in cloud-generated log data. The framework uses a modified version of the beetle antennae search optimization method to fine-tune parameters more precisely since classification success depends on accurate hyperparameter optimization. The improved XGBoost models were tested using real-world log datasets that are accessible to the public. The top models achieved accuracy rates over 99.98% in the experimental findings, which showed remarkable performance and highlighted the potential of the proposed method as a flexible and adaptive solution for anomaly identification in cloud logging systems. bookPart Springer, Cham 546 561 10.1007/978-3-032-30008-9_44 LNNS Lecture Notes in Networks and Systems: ICICC 2026: Proceedings of International Conference on Innovative Computing and Communications, volume 2037