Hybrid Whale Optimization based Bidirectional Gated Recurrent Unit with Pre-trained CNN model for Software Fault Detection

The deep learning model (DL) S-ResNet-152 (Squeeze-based ResNet-152) is used to pull out the features. Then, a bidirectional gated auto network (Bi-GRU-AN) is used to predict software faults based on a mixed DL model. A modified whale optimization algorithm and a crisscross optimization algorithm (MWOA-CS) are used to find the best weight for the suggested classifier. In MWOA-CS, the position of each dimension of the optimization problem is changed by a random better WOA optimization method while the repetitive process is going on. In this work, experiments are done on two software defect data sets (Kamei and PROMISE) to show how well the model classifies the data. The precision, recall, accuracy, and F1-score are the four complete evaluation measures used. The results show that the framework projected in this paper does a great job of classifying the two types of software fault data sets and is much better than the baseline models. then precision rate as 98.32 and then recall rate as 93.24 and then F-1 score as 94.53 correspondingly.