In modern networked systems, managing mean queuing delay is essential, particularly in environments where traffic is multi-class, bursty, and highly correlated. This study explores strategies for optimizing mean delay in queuing systems that handle diverse traffic classes with varying priorities and burstiness. Traditional delay control mechanisms often fail under conditions of irregular traffic, resulting in inefficiencies and increased latency. We propose a novel framework that integrates adaptive queuing policies with traffic prediction techniques to effectively reduce mean delay while maintaining service quality for all traffic classes. Simulation results demonstrate that our approach significantly reduces mean queuing delay across multi-class traffic scenarios compared to standard queuing models. This work has potential applications in high-demand environments such as telecommunications networks, cloud services, and data centers where efficient traffic handling is critical.

This study evaluated multiple machine learning models to predict customer lifetime value (CLV), including ensemble methods, linear regression, and deep learning architectures. Our comparative analysis demonstrated that ensemble models, such as Random Forest and Gradient Boosting, provided a robust balance of accuracy and interpretability, while deep learning approaches excelled in capturing complex data patterns but required higher computational resources. The results highlight the trade-offs businesses must consider in selecting models based on data availability, scalability, and interpretability. The insights from this research enable businesses to identify high-value customers effectively, optimize marketing strategies, and drive sustainable profitability.

Banking fraud has become a pervasive challenge, necessitating innovative solutions to protect financial institutions and their customers. This study investigates the effectiveness of supervised machine learning algorithms in detecting fraudulent activities within the banking sector. We conducted a comparative analysis of five widely used algorithms: Logistic Regression, Random Forest, Support Vector Machines, Gradient Boosting, and Neural Networks. Using a real-world banking dataset, we employed robust preprocessing and fine-tuning techniques to address class imbalances and optimize model performance. The evaluation metrics, including accuracy, precision, recall, F1-score, and area under the ROC curve (AUC), revealed that Gradient Boosting and Neural Networks consistently outperformed other models, achieving high precision and recall rates. The results highlight the potential of machine learning to detect subtle patterns of fraud while minimizing false positives and negatives. Furthermore, we discuss the implications of these findings for real-time fraud prevention systems and emphasize the importance of algorithm selection and scalability in operational environments.