Optimizing Soft Failure Detection in Elastic Optical Networks: A Comparative Analysis of Machine Learning Algorithms Note: Sub-Titles are not Captured in Xplore and Should not be Used

This document is a model and instructions for LATEX. Maintaining efficiency and performance in Elastic Optical Networks (EONs) requires a robust system for detecting “soft failures” -issues that gradually impact network performance without an immediate outage. A reliable detection system helps enable proactive maintenance, better use of resources, and fewer downtime and repair costs, all of which contribute to smooth and high-quality data transmission. This study explores machine learning techniques to detect soft failures in Elastic Optical Networks (EONs), a key element in supporting network reliability. We evaluated several classification algorithms, including random forest, gradient boost, support vector machine, and decision tree algorithms using simulated network data. To measure the effectiveness of each algorithm, we analyzed accuracy, precision, recall, and F1 score. Our results indicate that the Random Forest, Decision Tree, and Gradient Boosting algorithms all reached a perfect accuracy of 100%. However, Random Forest outperformed the others in precision, recall, and F1 score, showing it as the most effective overall. Although all three algorithms are suitable in terms of accuracy, Random Forest stands out due to its consistently superior results across all evaluated metrics. This comprehensive comparison highlights the value of machine learning in enhancing soft failure detection within EONs, with Random Forest emerging as a highly reliable choice.