Improving Software Quality Assurance Using a Cross-Context-Attentive Cosine Convolutional Model
Deena Sunil, Devika Dabke, Surendra Kumar Saini, Ravendra Ratan Singh Jandail, MWP Maduranga, S. N. Raw · 2025
The software quality determines system performance and reliability. As software systems are becoming increasingly complex, this is an important determinant. Detection of defects at the early stages of the development lifecycle improves software reliability and reduces costs. Traditional Software Defect Prediction (SDP) models suffer from class imbalance, complexity of datasets, and inefficient use of contextual information, which affect the accuracy of their predictions. Therefore, the development of reliable software systems still requires a more robust and efficient defect prediction model. In this manuscript, Improving Software Quality Assurance Using a Cross-Context-Attentive Cosine Convolutional Model (Sox-CNN-XA) is proposed. This study uses the NASA-MDP dataset containing historical records of software defects and various software metrics. Data pre-processing included data cleaning, normalization, and splitting, which was done in order to enhance the quality of the data and make it ready for training and validation purposes. A new framework for Software Defect Prediction (SDP) has been proposed by using the integration of Cosine Convolutional Neural Networks (Cos-CNN) and Cross-Contextual Attention (XCA), which captures spatial and contextual features of data effectively. The hyperparameters in this model have been fine-tuned with Sardine Optimization Algorithm (SOA) in order to get improved performances over diverse datasets. Experimental results show that the proposed framework achieves an accuracy of 99.9%, a true positive rate of 99.9%, and a precision of 99.8% while maintaining a false positive rate of 7.2%. The results clearly prove that the model is performing better than traditional methods. The conclusion shows that the framework can increase the reliability of defect prediction, thus enabling software practitioners to deliver high-quality and maintainable software within tight deadlines and budgets