African buffalo optimized multinomial softmax regression based convolutional deep neural network for software fault prediction

Abstract

Software fault prediction is an essential part of the software quality assurance to detect faulty software modules depending on software measurement data. This models area of the research generates software engineers focus to development activities on fault-prone code for increasing the software quality. Many existing designed for improving the software quality by various methods. The software failure cause prediction accuracy and defect detection rate of the conventional techniques were not sufficient. Introduced an African Buffalo Optimized Multinomial Softmax Regression based Convolutional Deep Neural Learning (ABOMSR-CDNL). ABOMSR-CDNL model to enhance the software reliability through predicting root cause of software failure at earlier stage. ABOMSR-CDNL Model comprises four layers, namely input layer, two hidden layers and output layer and software program codes, event log files as an input layer then transmits the software program codes to the hidden layer 1. Construct the projects portfolio with help of optimal parameters selected from event log files then sent to the hidden layer 2. This method reduce the amount of time taken for examining the failure behaviour of system application. ABOMSR-CDNL Model applied multinomial softmax regression analysis in hidden layer 2 this objective of determining the cause of higher accuracy the result is sent to the output layer. Result illustrates ABOMSR-CDNL Model is increasing the accuracy, false positive rate and minimized the software failure identification time.