Theories to maximise influence in social networks have been investigated vastly by researchers in recent times. Various strategies have been proposed to identify initial seed node set to diffuse the information in the network. In this manuscript, a seed node selection strategy has been proposed based on the clustering coefficient and betweenness centrality scores of the nodes in the graph (CLUS-BET). The influence spread (in the underlying network) is measured using Independent Cascade Model (ICM) by selecting top k seed nodes having high CLUS-BET score. The network coverage obtained are \(72.48\%\) , \(65.85\%\) , \(51.12\%\) and \(53.15\%\) for CORA, Citeseer, PubMed and Amazon Computers network respectively. The results obtained are better in comparison to any of the classical centrality algorithms’-based seed node selection. Also, the results are compared with a recent state of art method GRAIN based seed selection strategy. The manuscript further testifies that the quality of nodes selected based on CLUS-BET property improves the Graph Convolution Network (GCN) based node classification accuracy. CLUS-BET based GCN produces a node classification accuracy of \(90\%\) , \(77\%\) and \(89.2\%\) for CORA, Citeseer and PubMed networks respectively. The results surpass the classification accuracy of base GCN model and its variations in contemporary state of art. Also, the property helps in identifying the most prominent influential clusters or groups of the network responsible for spreading the information in the network to a vast majority. The analytical explanation suggests that these are the groups having most number of nodes with high CLUS-BET score. The proposed idea is scalable with the graphical network as it avoids Monte Carlo simulation rounds with Greedy method for information cascade to identify seed nodes. The metric calculation for CLUS-BET is simple, less time consuming and effective.
