.png){kind=icon}
.png){kind=icon}
Return
Software failure prediction using hybrid deep learning model with optimization-enabled feature selection
.png){kind=icon}
PRE
AI
DOI：10.1080/03610918.2025.2571974.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
BackgroundTo recognize faults, Software Fault Prediction (SFP) approaches are employed at the early phases of software development life cycle (SDLC). Nowadays, Software Defect Prediction (SDP) is gaining popularity, which is often executed, frequently with the employment of Machine learning approaches. Nonetheless, current Machine learning-based techniques require manually extracted features that can be time-consuming, laborious, and frequently not successful in capturing semantic information contained in defect tracking system. Deep Learning (DL) approaches present experts with capability to automatically retrieve high-dimensional data.ObjectivesThus, an advanced hybrid deep learning model is proposed in this research for SFP. Primarily, input data is fed to pre-processing stage, executed by min-max normalization. Subsequently, feature selection is done using Double Exponential-Coati optimization approach (DE-COA) that is proposed by combining Double Exponential Smoothing (DES) and Coati optimization algorithm (COA). Subsequently, SFP is done by EfficientNetB0-ResNet50, which is proposed by the combination of EfficientNetB0 and ResNet50 models.ResultsAt last, analysis is done by employing accuracy, sensitivity and specificity metrics, whereas DE-COA attained enhanced accuracy of 0.968, better sensitivity of 0.990, and better specificity of 0.873.
Keywords:
Deep learning
Feature selection
Optimization
Software development
Software failure prediction
Journal
C
IF:
0.8
Papers:
213
Citations:
4.7K