Machine Learning Classification and Metaheuristic Optimization for Resilient Instant Messaging Networks

Abstract Nowadays, Enterprise instant messaging systems face serious dynamic workloads, resource constraints, and service degradation that can affect communication reliability. This paper proposes an AI-enabled framework that integrates machine learning classification with ADMM-LASSO optimization for resilient messaging management. Logistic Regression, Decision Tree, Random Forest, Gaussian Naive Bayes, Support Vector Machine, and Stochastic Gradient Descent are employed to classify operational states from system telemetry. ADMM-LASSO then performs sparse feature selection and resource optimization to support workload balancing and adaptive resource allocation. Experimental results demonstrate effective classification of normal and high-load states, stable optimization convergence, and improved resource-management efficiency. The findings show that combining machine learning and sparse optimization provides a scalable approach to intelligent and resilient management of enterprise instant messaging systems.