Graphs are mathematical structures used to model relationships and interactions in diverse fields, including social networks, biology, and industrial applications. In sectors like banking and telecommunications, the exponential growth of data has highlighted the need for efficient graph analysis methods. This paper focuses on the enumeration of SubGraphs of Interest (SGI) within large multigraphs to enhance fraud detection in an industrial context. We present a novel approach that leverages pruning functions to remove less relevant nodes and edges, thereby constructing informative connected components. Using synthetic datasets, we evaluated the performance of the proposed approach and assessed the impact of node and edge classifications on SGI enumeration. We also explore the use of machine learning classifiers, including XGBoost and Random Forest Classifier, to enhance classification performances.