Abstract:
Signed and unsigned networks are types of complex networks that represent relationships between nodes in a graph, Edges in unsigned networks are unweighted, indicating that links are either present or absent. In contrast, signed networks have weighted edges that can be positive or negative to signify positive or negative connections. Nodes are classified into specified classes based on their structural characteristics and other factors, a process known as node classification. Node classification gets more difficult in directed signed networks where relationships are asymmetric and might be either positive or negative therefore conventional techniques for classification in undirected networks fails in this situation. However, because of its importance to social networks, trust prediction, and online review systems, there has been an increase in interest in this field of study. Node classification in directed weighted graph incorporating graph features i.e. NCDWG (Graph features) look into the problem of node classification in directed signed networks and provide an approach to enhance the performance of classification.Feature extraction techniques; Deep Walk and node2vec, are used for feature extraction. To enhance the feature extraction process node degree, average weights and weights of the nodes are also considered as an additional feature for downstream classification task. There has been a rise in the creation of novel node classification methods for directed signed networks. To enhance the classification performance, our suggested method solves this problem by combining node degree, average weights and weights of the nodes as an additional feature along with deep learning-based methods.
