Software defect prediction using multi-source feature gating fusion
With the increasing scale and complexity of current software development,improving software quality while ensuring efficiency has become a focus and challenge in the field of software engineering.Software defect prediction is an important re-search direction in software quality assurance,aiming to help software practitioners anticipate potential defective modules in soft-ware products and allocate testing more effectively.Previous studies have mainly extracted software features to build defect pre-diction models,but typically only utilized single-type features as model inputs,and lacked effective feature integration,leading to the need for improvement in defect prediction performance.This paper proposed a software defect prediction method based on multi-source feature gating fusion(DP-GM).Firstly,abstract syntax trees and word embedding models were utilized to obtain code semantic representations.Then,gated recurrent unit(GRU)was employed to extract features from semantic feature vectors and traditional feature vectors.Finally,a gating mechanism was used to fuse multi-source features to train the model and perform software defect prediction.Experimental results showed that compared to three representative baseline methods,the proposed method improved recall and F1 value by 35.3％and 10.5％respectively.Therefore,the method proposed in this paper could enhance the accuracy of software defect prediction and help software practitioners improve development efficiency.
NETL
NSTL
万方数据
