Fault Prediction Using Deep Neural Network

On the use of deep learning in software defect prediction

“…Semantic features: Traditional features such as source code metrics are not sensitive to the semantics of programs, and can lead to poor prediction performance (Wang et al, 2016;Li et al, 2017;Huo et al, 2018;Cai et al, 2019;Chen et al, 2019;Dam et al, 2019;Fan et al, 2019a;Fan et al, 2019b;Liang et al, 2019;Qiu et al, 2019c;Xu et al, 2019;Zhang et al, 2020a;Bahaweres et al, 2021, Chatterjee et al, 2021Xu et al 2021a;Zhang et al, 2018;Zhang et al, 2021a). A wide range of solutions have been offered to tackle this.…”

“…Traditional features (such as lines of code, operand and operator counts, number of methods in a class, the position of a class in inheritance tree, and McCabe complexity) are not sensitive to programs' semantic information and hence harm defect prediction performance Li et al, 2017;Huo et al, 2018;Cai et al, 2019;Chen et al, 2019;Fan et al, 2019a;Fan et al, 2019b;Liang et al, 2019;Qiu et al, 2019c;Xu et al, 2019;Zhang et al, 2020a;Bahaweres et al, 2021, Chatterjee et al, 2021Xu et al 2021a;Zhang et al, 2018;Zhang et al, 2021a). Wang et al, 2016 and…”