Research on Efficient Software Defect Prediction Using Deep Learning Approaches

Software Defect prediction results provide a list of source code artifacts that are prone to defects. Quality assurance teams can effectively devote more energy and allocate limited resources to defect-prone source code verification software products. A module that identifies defect prediction methods for frequent defects before the start of the testing phase. Measurement-based defect-prone modules improve software quality and reduce costs, leading to effective resource allocation. The previous method doesn't analyze the defect pattern, and it has less performance during software development. This work introduces a deep learning-based Pattern-based Modified Hidden Markova Fault Tree (PMHMFT) framework to extract the hidden fault analysis during cross-project validation. The proposed Modified Hidden Markova Fault Tree algorithm constructs the defect fault tree to analyze the cross-project code defect. To prevent defect based on software metrics software prediction model are used. Hidden Markova Fault tree-based classification categorize component as defective and non-defective. Using a Levy flight, optimize the method to search the fault classes efficiently compared to another method. The Markova Fault Tree model construct fault tree based given data; it is easy to identify the fault in software platform. The proposed PMHMFT to implement evaluate the performance using k-fold validation. Thus, the proposed work on software defect prediction achieves higher accuracy in true classification and prediction with less error rate. The software defects are predicted, and these predicted defects are optimized by using Levy flight optimization. Our proposed PMHMFT technique is very useful technique for predicting software defect and gives the better prediction rates in effective manner.