Finding the interaction flow between various compounds found in chemicals or genes as graph interactions is made possible by the important research field known as graph pattern mining. The task of mining the graphs to extract the interaction flow would be more challenging. The most common data mining method, called clustering, groups the nodes that make up graphs with extensive communication to discover the graphical interaction flow. The process of graph pattern mining can be carried out using a variety of research methodologies. We introduced Flow-Based Algorithms (FBA) for Local Graph Clustering in our previous research approach, which aimed to group various graphs based on flow. However, the presence of numerous irrelevant features, which needed to be avoided for better performance, prevented this method from increasing the clustering accuracy. This issue introduces the Irrelevant Feature aware NMF clustering method, which focuses on the targeted research work (IF-NMFCM). To prevent other associated dataset complexities and increase the cluster accuracy of strategies during this work, it is crucial to remove irrelevant features. Here, principal component analysis is initially used for data pre-processing. Then, using the PSO method or the particle swarm optimization method, feature selection is carried out in order to avoid the irrelevant features. Adopting this approach would produce the ideal set of features, which could increase the accuracy of clustering. Finally, pre-processed data would be represented using a graph after being clustered using the hierarchical NMF clustering method. The proposed research work leads to provide higher results than the current method in terms of improved accuracy rate, according to the overall analysis of the proposed research method done in a Mat lab simulation environment.

Graph mining, Clustering, Pattern mining, Useful information extraction.