Artificial Intelligence-Based Software Engineering Framework for Automated Program Quality Assessment

Abstract

The evaluation of the quality of software products plays an important role in software engineering, enabling the creation of robust, maintainable, and secure software applications with great performance levels. With the growth in size and complexity of modern software projects, traditional methods for quality assessment using hand-crafted software metrics and supervised machine learning need much feature engineering and labeled defect data. In order to overcome difficulties outlined above, in this research, framework for automated program quality assessment based on Artificial Intelligence-Based Software Engineering (AISE) is suggested. This framework encompasses use of AI-SEPQA dataset, data preprocessing, feature extraction through CodeBERT (Bidirectional Encoder Representations from Transformers), quality prediction using hybrid approach Dynamic Tunicate Swarm Algorithm-fused Least Square Support Vector Machine (DTSA-LS-SVM). AI-Based Software Engineering Program Quality Assessment (AI-SEPQA) dataset consisting of 12,000 software modules is used and preprocessed via comment removal, lexical normalization, and tokenization. Machine Learning-based DTSA-LS-SVM quality prediction for evaluation of various software quality metrics such as defect proneness, code complexity, maintainability, reliability, code standards conformance. DTSA fine-tunes LS-SVM parameters to ensure convergence and generalization, while LS-SVM helps in predicting software quality metrics with better classification accuracy and prediction. Experimental evaluation is done using Python, and proposed framework is evaluated on the basis of various performance measures, including accuracy of 98.25%. The results obtained by applying the AISE framework have been found to be statistically significantly better than those obtained through the use of traditional approaches (p = 0.003). The framework offers a smart way of assessing program quality automatically.