HECS: A Hypergraph Learning-Based System for Detecting Extract Class Refactoring Opportunities

HECS is an advanced tool designed for Extract Class refactoring by leveraging hypergraph learning to model complex dependencies within large classes. Unlike traditional tools that rely on direct one-to-one dependency graphs, HECS uses intra-class dependency hypergraphs to capture one-to-many relationships. This allows HECS to provide more accurate and relevant refactoring suggestions. The tool constructs hypergraphs for each target class, attributes nodes using a pre-trained code model, and trains an enhanced hypergraph neural network. Coupled with a large language model, HECS delivers practical refactoring suggestions. In evaluations on large-scale and real-world datasets, HECS achieved a 38.5% increase in precision, 9.7% in recall, and 44.4% in f1-measure compared to JDeodorant, SSECS, and LLMRefactor. These improvements make HECS a valuable tool for developers, offering practical insights and enhancing existing refactoring techniques.