Software Defect Prediction Harnessing on Multi 1-Dimensional Convolutional Neural Network Structure

Developing successful software with no defects is one of the main goals projects. In order to provide a project anticipated quality, prediction plays vital role. Machine learning, and particularly deep have been advocated for predicting defects, however both suffer from inadequate accuracy, overfitting, complicated structure. this paper, we aim address such issues in defects. We propose novel structure 1-Dimensional Convolutional Neural Network (1D-CNN), learning architecture extract useful knowledge, identifying modelling knowledge data sequence, reduce finally, predict whether units code are prone. design large-scale empirical studies reveal proposed model's effectiveness by comparing four established traditional machine baseline models state-of-the-art baselines defect based on NASA datasets. The experimental results demonstrate that terms f-measure, an optimal modest 1D-CNN dropout layer outperforms 66.79% 23.88%, respectively, ways minimize overfitting improving performance According results, seems be may applied adopted practical problem engineering. This, turn, could lead saving development resources producing more reliable software.