Hyperparameter optimization using Deep learning methodology for Software bug prediction and test Case prioritization to improve Software quality

Abstract

Software defect is known as any flaw, mistake, bug, omission, newlinefailure, or fault in software that can cause computers to produce wrong newlineor unexpected results. The cost and duration needed to create a highquality newlinesoftware product are always rising as a result of software flaws. newlineTo deliver high-quality software, the product should contain as few newlinefaults as feasible. Specifically, automated software testing can be used to newlineeliminate manual labor, shorten testing duration, and improve testing newlineperformance. One of the most time-consuming and expensive software newlineactivities is locating and repairing software bugs. A technique for newlinelocating faulty software components is called Software Defects newlinePrediction (SDP). It s understood from the literature survey there is no newlineproper sequence planning and designing of test cases. newlineProposed work implement an appropriate test sequence by newlineusing application module dependencies for deterministic algorithm. newlineAdvanced Deep Learning (DL) models may be employed to classify newlinedefective software components. So hyperparameters must be optimized newlineusing metaheuristic methods because they have a significant impact on newlinehow well any DL model performs. Then perform prioritization of test newlinecase using DDPF model. newlineThe foremost contribution is that to ensure software newlinedependability and trustworthiness, the proposed method involves newlinemodule sequencing using real time data. Analyze the application newlinevi newlinedependency matrix structure algorithms in order to effectively plan for newlinetest sequencing during planning phase and perform optimization. newlineSecond contribution provides an effective, one-of-a-kind newlineMetaheuristic Optimization with Deep Learning-based SBP (MODLSBP) newlinetechnique. This method involves developing a hybrid Convolution newlineNeural Network (CNN) bi-directional long short-term memory newline(BiLSTM) in order to forecast software issues. Chaotic Quantum newlineGrasshopper Optimization Algorithm (CQGOA) is utilized to optimize newlinethe hyperparameters of CNN-BiLSTM models, hence improving their newlinepredictive accuracy. To validate the higher presentation