AI-Enabled Edge Intelligence for Real-Time IoT Analytics A Distributed Framework for Low-Latency Decision-Making and Resource Optimization

Abstract

The study investigated an AI-enabled distributed edge-intelligence framework for real-time Internet-of-Things (IoT) analytics focused on low-latency decision-making, resource optimisation, energy efficiency and quality of service (QoS) performance. In study, a dataset of 2,400 observations and 22 variables that characterize different IoT workloads, network characteristics, computational resource usage, inference properties, offloading behavior, and operational outcomes, and adopted a computational evaluation design. The analysis performed these comparisons using descriptive statistics, independent-samples t-tests, one-way ANOVA, correlation analysis, multiple linear regression, and five supervised machine-learning models. The AI-adaptive strategy not only significantly cut the end-to-end latency down from 78.00ms to 52.63ms, it also lowered energy consumption, utilization of CPU and memory, packet loss, queue delay and inference time. The resource efficiency improved from 46.78 to 59.49, and the decision correctness improved from 26.4% to 52.6%. The number of QoS violations dropped significantly from 49.4% with static processing to 11.6% with AI-adaptive orchestration. Orchestration mode, queue delay, inference time and network round-trip time (RTT) were identified as major factors that affect latency through regression analysis. Logistic Regression had the best ROC-AUC score of 0.639 among all classifiers tested. The results highlight the potential of leveraging AI-driven edge orchestration to enhance responsiveness, efficiency, and reliability in distributed IoT setups overall.