Efficient log monitoring is essential for ensuring the performance and reliability of scalable cloud computing systems. As these systems grow in complexity and scale, traditional monitoring approaches become less effective and more resource-intensive. This work explores the integration of natural language processing (NLP) techniques with advanced classification algorithms to identify anomalies in cloud-based log data. Given that classifier accuracy heavily relies on optimal hyperparameter configurations, a metaheuristic approach is employed to automate and refine the tuning process. A novel variant of the artificial bee colony (ABC) algorithm, named self-adaptive ABC (SAABC) is proposed and benchmarked using publicly available real-world datasets. Experimental results indicate strong performance, with the SAABC achieving an accuracy of .985275, demonstrating its potential for intelligent and scalable log anomaly detection.