Presentation of a new method for predicting software defect using neural network combination and grasshopper algorithm

• Published in: مجله مدل سازی در مهندسی Vol. 17; no. 57; pp. 201 - 214
• Main Authors: somaye shabani zade rabori, vahid khatibi bardsiri, amid khatibi bardsiri
• Format: Journal Article
• Language: Persian
• Published: Semnan University 01.06.2019
• Subjects: grasshopper algorithm; perceptron neural networks; reliability; software defect prediction
• ISSN: 2008-4854; 2783-2538
• DOI: 10.22075/jme.2019.15226.1514

Software development cycle includes analysis, design, implementation and testing, and several other phases. The software testing phase is one of the costly stages of software development, and should be effectively implemented so that the software can be accessed without error by the users. One of the most effective activities for software development and its reliability is to predict software flaws before reaching the testing stage, which helps to save time in the production, maintenance and cost of software. One of the most effective models for predicting software flaws is the use of multilevel perceptron neural networks with post-error training algorithm. One of the weaknesses of the post-error training algorithm is the possibility of trapping the neural network at the local minimum points. Considering the potential of hyper-algorithms in exiting the local minimum mines and finding the minimum in general, in this paper, the combination of grasshopper meta-heuristic algorithm and post-error training algorithm were used to solve the weakness of the training algorithm and to improve its accuracy in predicting software defect. Is. In order to evaluate the results of the proposed model, neither the actual database was used nor the cross-evaluation method was the basis for presenting the results. The proposed model's performance has been compared with six defective software prediction models. The results of this comparison show that the proposed model is able to provide more accuracy and accuracy in comparison to other models in a large number of data sets.