Abstract The rapid expansion of graph-structured data necessitates advanced analytical methods for understanding complex organizational communication networks. This study presents a comprehensive graph representation learning framework tailored for large scale organizational email networks, integrating deep graph neural networks (GNNs), graph autoencoders, and explainable artificial intelligence (XAI). Utilizing the ia-email- EU-dir dataset, comprising over 265,000 nodes and 420,000 directed edges, we systematically evaluate classical embedding techniques against... The rapid expansion of graph-structured data necessitates advanced analytical methods for understanding complex organizational communication networks. This study presents a comprehensive graph representation learning framework tailored for large scale organizational email networks, integrating deep graph neural networks (GNNs), graph autoencoders, and explainable artificial intelligence (XAI). Utilizing the ia-email- EU-dir dataset, comprising over 265,000 nodes and 420,000 directed edges, we systematically evaluate classical embedding techniques against advanced GNN architectures. Our proposed hierarchical encoderdecoder architecture, specifically the Adversarially Regularized Graph Autoencoder (ARGA), demonstrates superior performance in capturing both local connectivity and global topological dependencies. ARGA achieved the highest predictive accuracy for link prediction tasks, yielding an Area Under the Curve (AUC) of 0.948 and an Average Precision (AP) of 0.935, alongside the lowest reconstruction loss. Furthermore, the framework incorporates multiple XAI modules including GNNExplainer, GraphSHAP, and attention weight visualization to decode latent representations, thereby identifying critical communication hubs, influential subgraphs, and hidden community structures. By bridging scalable representation learning with model interpretability, this unified pipeline offers a robust solution for organizational analytics, enabling practical applications such as anomaly detection, knowledge flow mapping, and hierarchy discovery. Ultimately, the empirical findings indicate the profound efficacy of adversarially regularized probabilistic models in deciphering the intricate, evolving dynamics of sparse, scale free enterprise communication networks, paving the way for trustworthy AI in enterprise decision support and organizational behavior analysis. Read More Read Less

Abstract The evolution of network management toward user centric Quality of Experience (QoE) necessitates intelligent routing frameworks capable of balancing multimedia quality, communication reliability, and energy efficiency in dynamic environments such as fifth generation Vehicular Ad Hoc Networks (5G VANETs). This study presents an explainable machine learning approach to analyze and predict QoE-aware energy-efficient routing decisions using a simulated dataset of 10,000 routing instances encompassing... The evolution of network management toward user centric Quality of Experience (QoE) necessitates intelligent routing frameworks capable of balancing multimedia quality, communication reliability, and energy efficiency in dynamic environments such as fifth generation Vehicular Ad Hoc Networks (5G VANETs). This study presents an explainable machine learning approach to analyze and predict QoE-aware energy-efficient routing decisions using a simulated dataset of 10,000 routing instances encompassing network, multimedia, and energy related parameters. A Random Forest classifier with balanced class weighting was employed, achieving near perfect predictive performance with 99.9% accuracy, 99.56% balanced accuracy, and a Matthews Correlation Coefficient of 0.998. To address the "black box" nature of machine learning, SHapley Additive exPlanations (SHAP) provided both global and local interpretability, consistently identifying packet delay and video quality (PSNR) as the dominant determinants of routing outcomes. High_QoE_Route instances exhibited superior performance, characterized by lower delays (32.94 ms) and higher PSNR (43.96 dB), while maintaining competitive energy consumption profiles. Statistical analysis confirmed significant differences across routing classes for delay and PSNR. The findings demonstrate that effective routing emerges from optimizing multiple performance dimensions simultaneously, where delay and perceptual quality outweigh packet loss in influencing user satisfaction. By integrating high predictive accuracy with model transparency, the proposed framework enhances trustworthiness and practical deployability, contributing toward more reliable, sustainable, and intelligent vehicular communication systems for next generation transportation infrastructures. Read More Read Less

Abstract The rapid evolution of the Internet of Vehicles (IoV) introduces significant challenges for dynamic routing due to high mobility and complex interdependencies among network parameters. Datasets used for machine learning based routing decisions frequently exhibit multicollinearity, leading to unstable model estimates, inflated variances, and reduced interpretability. Traditional mitigation techniques, such as Principal Component Analysis, often sacrifice valuable domain specific information by transforming or eliminating variables.... The rapid evolution of the Internet of Vehicles (IoV) introduces significant challenges for dynamic routing due to high mobility and complex interdependencies among network parameters. Datasets used for machine learning based routing decisions frequently exhibit multicollinearity, leading to unstable model estimates, inflated variances, and reduced interpretability. Traditional mitigation techniques, such as Principal Component Analysis, often sacrifice valuable domain specific information by transforming or eliminating variables. This study proposes a comprehensive framework combining extensive multicollinearity diagnostics with a Ridge Classifier to enhance routing decision making in dynamic IoV networks. Using a synthetic dataset of 4,000 instances comprising 28 numerical features, we performed rigorous diagnostics including Variance Inflation Factor (VIF), condition indices, and Belsley analysis confirming severe feature interdependencies. Instead of discarding variables, we applied L2 regularization via a Ridge Classifier to stabilize coefficient estimates while preserving all domain relevant predictors. The optimized Ridge model achieved a test accuracy of 84.63% and a 5-fold cross validation accuracy of 84.50%, significantly outperforming a baseline Random Forest classifier at 81.13%. Additionally, implementing class weights successfully improved minority class recall, highlighting the framework's adaptability to imbalanced routing scenarios. Ultimately, the empirical results Proved that ridge regularization effectively mitigates numerical instability caused by multicollinearity without sacrificing critical vehicular network metrics. This research provides a reliable, interpretable, and computationally efficient analytical foundation for intelligent routing protocols, contributing to more resilient next generation intelligent transportation systems. Future work will explore real world deployment and advanced class balancing strategies. Read More Read Less