Enhancing Software Quality Prediction through Source Code Analysis with the Firefly Algorithm

Abstract

For software systems to operate reliably and effectively, software quality assurance is essential. Conventional techniques for predicting software quality frequently rely on manual code inspection and testing, which can be laborious and prone to errors. This study suggests a unique method for predicting software quality assurance through deep learning and data mining analysis of source code. Developing an automated system that forecasts software quality using source code analysis is the primary goal of the project. In order to identify significant patterns and characteristics from the source code and capture both structural and semantic information, the suggested system makes use of data mining techniques. A type of deep learning model employed to understand the intricate connections between software and the extracted features is the convolutional neural network (CNN). A large collection of source code samples and related quality metrics will be gathered in order to achieve this goal. The source code samples will be used to extract several code metrics, including code complexity, code duplication, and code smells. The data mining and deep learning models will use these metrics as input features. Pre-processing will be applied to the gathered dataset to address any noise or inconsistent data. The most pertinent and instructive elements for software quality prediction will be found using feature selection and dimensionality reduction approaches. Using the quality metrics and extracted features, deep learning algorithms will be developed and optimized using the training set. The models will undergo optimization processes, including hyper parameter tuning and regularization, to achieve optimal performance and generalization capabilities. The trained models will be evaluated using the validation set, fine- tuning them if necessary.