An Enhanced Patch Optimization Technique for MultiChunk Bugs in Automated Program Repair

Automated program repair techniques leveraging deep learning have shown remarkable performances in bugrepair. These techniques commonly employ pretrained neural machine translation (NMT) models to generatepatches for a buggy part of the source code. However, when dealing with multiple buggy code chunks in variouslocations, current methods face challenges in effectively selecting and combining these patches for optimalrepair. This paper identifies limitations within one of the existing methods used for optimizing patches relatedto multiple buggy code chunks and proposes an enhanced patch optimization technique to address theseshortcomings. The primary aim of this study is to improve the process of selecting and combining patchesgenerated for groups of buggy chunks. Through experiments conducted on a dataset, this paper demonstratesthe efficacy of the proposed patch optimization technique, showcasing its potential to enhance the overall bugrepair process. This study highlights the importance of patch optimization in bug repair by addressinglimitations and enhancing the repair process. KCI Citation Count: 0