The influential nodes in complex network are the key of high effective information spreading. Several techniques have been developed for the discovery of such nodes, including centrality-based approaches, machine learning-based approaches, and deep learning-based approaches. This paper proposes CNNG, a novel hybrid deep learning model combining Convolutional Neural Networks (CNN) and Graph Attention Networks (GAT) for predicting node influence. CNNG processes each node’s local subgraph through parallel CNN and GAT components to capture both structural patterns and relational dependencies. The model is trained in a supervised regression setting, using node influence scores generated via SIR simulations. Experiments on twelve real networks show that CNNG achieves the best average Kendall’s Tau coefficient of 0.7510, outperforming the second-best method by 2.85% as well as the highest average Monotonicity Index of 0.9954 and the highest average Jaccard Similarity of 0.7298 over ten networks. Further, CNNG provides these results in a reasonable amount of computing time, which demonstrates the practicability, efficiency and generality of CNNG for the complex network analysis.

Identifying influential nodes in complex networks is a critical task with significant applications across various domains. Traditional methods often fail to address the varied nature of node influence, particularly in large-scale networks. To overcome these limitations, this paper presents a machine learning-based method for identifying influential nodes using robust feature engineering and prediction modeling. Specifically, feature vectors are constructed using Extended Degree Centrality (EDC), Eigenvector Centrality (EC), and a composite metric that integrates EDC and VoteRank using a weighting parameter α together with EC to capture both local and global structural properties of the network. In order to achieve representativeness and scalability, a systematic sampling is proposed and utilized to ensure that nodes with diverse struc tural and influential properties are included in the training set. Then Random Forest Regressor model employed to predict node influentiality across diverse network datasets. Experimental results on nine real-world network datasets demonstrate the method’s superiority over baseline methods (DC, KS, KSIF, CC, EC, VR, BC, HI, MDD, CR, PR, and RCNN), achiev ing the highest average Kendall Tau of 0.8158, outperforming the second-best methods by 13.9% over EC and 17.49% over RCNN, and improvements in Jaccard similarity by 13.95% over EC and 8.47% over RCNN. The method’s robustness across network sizes and structures further confirms its scalability and effectiveness.

Identifying influential nodes in complex networks is a critical task with a wide range of applications across different domains. However, existing approaches often face trade-offs between accuracy and computational efficiency. To address these challenges, we propose 1D-CGS, a lightweight hybrid model that integrates the speed of one-dimensional convolutional neural networks (1D-CNN) with the topological representation power of GraphSAGE for efficient node ranking. The model employs a lightweight input representation based on two straightforward and significant topological features: node degree and average neighbor degree. These features are processed through 1D convolutions to extract local patterns, followed by GraphSAGE layers to aggregate neighborhood information. We formulate node ranking as a regression problem and use the Susceptible–Infected–Recovered (SIR) model to generate ground truth influence scores. 1D-CGS is initially trained on synthetic Barabási–Albert networks and then applied to real-world networks. Experimental evaluations on twelve real-world networks demonstrate that 1D-CGS significantly outperforms traditional centrality methods and recent deep learning models in ranking accuracy, while maintaining very fast runtime. On these networks, the proposed model achieves average improvements of 4.73% and 7.67% over the best-performing deep learning baselines. It also achieves an average Monotonicity Index (MI) score of 0.9955, indicating highly unique and discriminative rankings. Furthermore, additional comparisons on benchmark networks show that 1D-CGS achieves the highest average Kendall’s Tau of 0.8165 among the evaluated approaches, confirming its strong generalization capability. Overall, the results demonstrate that 1D-CGS is both efficient and scalable, making it suitable for large-scale applications.