An explainable Grey Wolf optimized extreme learning machine framework for modulation classification in cloud environment

Abstract

Abstract Automatic modulation classification (AMC) is a key component in modern wireless communication systems, supporting efficient spectrum utilization and reliable data transmission. This paper presents a novel AMC model named as XGEM-Net, an explainable grey wolf optimized extreme learning machine model designed for cloud-centric environment. In the proposed model, the features have been extracted from pre-trained models like InceptionV3, ResNet-50, and MobileNetV2 and concatenated to form a unified feature vector. The combined features are input into an ELM optimized via Grey Wolf Optimization (GWO), forming the GWO-ELM model to classify various modulation types. Furthermore, the framework integrates explainable artificial intelligence using local interpretable model agnostic explanations (LIME) to deliver clear, instance level interpretability of classifier decisions. The proposed model has been evaluated in both standalone and cloud-based environments, including configurations with vCPU-4 (16 GB RAM), vCPU-8 (32 GB RAM), and vCPU-16 (64 GB RAM). Experimental results indicate that the model performs significantly better in the cloud setting, with the vCPU-16 64 GB configuration achieving 95.16% accuracy, 90.78% sensitivity, and 89.83% specificity. These findings demonstrate that the proposed approach consistently outperforms existing state-of-the-art methods in terms of classification accuracy.