Deep learning-based approach to predict software faults
Seema Kalonia, Amrita Upadhyay · 2024
Today&s;s software engineering is changing very fast, hence predicting software faults becomes crucial in enhancing software reliability and decreasing maintenance expenses. Although traditional fault prediction methods are partially effective, they have difficulties with the complexity and sizes of modern software systems. This complex behavior can be recognized by deep learning, which can learn complicated designs and associations in data. This chapter examines application of DL techniques such as convolutional neural networks (CNNs), recurrent neural networks (RNN), and long short-term memory network (LSTM), among other NN for predicting software faults. The effectiveness of these models in identifying modules prone to faults was demonstrated using the Metrics Data Program (MDP) datasets provided by NASA, which offer a rich resource of software metrics and fault information. Deep learning has become a revolutionary technology over recent years in several domains like natural language processing, computer vision, healthcare, etc. The chapter explored how deep learning can help predict faults in software programs, including its benefits, challenges, and opportunities ahead. The study also uses NASA&s;s MDP datasets for applying deep learning techniques to predict software defects. To enhance model performance, the RNNs, CNNs, and many other NN are used with NASA MDP datasets. The CNN and LSTM models achieve higher precision and accuracy than other deep learning methods, demonstrating their effectiveness in identifying fault-prone software modules and improving reliability.