Bug prediction based on deep neural network with reptile search optimization to enhance software reliability.

Software reliability is a far more important factor that influences quality of the software. Software bug identification is a critical aspect of software development method. Software reliability is greatly impacted by the existence of defects in the software, so must anticipate bugs in software. Nevertheless, software bug detection may be insufficiently accurate for practical application, and wide benefits of various may be used. To address these concerns, Modified Deep Neural Network (MDNN) is suggested for predicting software defects. Initially, raw data's are collected and pre-processed by using min–max normalisation that reorganises the data as in database. Then, utilizing principal component evaluation for reduce the dimensionality of pre-processed data. After reducing dimension select appropriate features using correlation based Fuzzy C-means Clustering Method (FCM). First, unnecessary characteristics are removed employing FCM, and then non-redundant features were extracted from every cluster utilising correlations value. After that selected features are given as an input for MDNN. MDNN is developed through optimal selection of weight parameter using Reptile Search Optimization (RSA) algorithm providing error as fitness. Finally, classifier predict bugs in software module which is used to improve software reliability performance is achieved. According to the simulation study, the proposed method achieves 98% accuracy, 0.02% error, 95% specificity, 90% recall, and 95% precision. This indicates that the proposed approach performs better than all prior options. Based on this proposed classification bugs are predicted and software reliability performance is improved. [ABSTRACT FROM AUTHOR]