Research On Software Defect Prediction Method Based On ICS-ANN

Software defects is the main reason for software running error.How to reduce the software defects has always been a problem to be solved to ensure the software quality.The purpose of static software defect prediction is find out the defect-prone software modules in new software versions so as to help quality assurance testing team focus on the defect-prone modules,improve the efficiency of software testing and find out more software defects.An increasing number of scholars use Artificial Neural Network(ANN)algorithm to build software defect prediction model.But the artificial neural network algorithm has more weights,the global optimum weight cannot be easily found by the conventional method such as gradient descent;in addition,the performance of prediction model decreases by the noise attributes and class imbalances existing in the software defect prediction datasets.To solve these problems,this paper's main work is as follows:Firstly,this paper applies the cuckoo search(CS)algorithm to the software defect prediction,and propose a software defect prediction method based on CS-ANN.The noise attributes existing in data are processed by the Correlation-based Feature Selection algorithm in this method;the weights of the artificial neural network are found by the cuckoo search algorithm;maximizing the value of F1-Measure as a standard is to determine the optimal threshold.The simulation experiment is performed through the Metrics Data Program(MDP)datasets that was published by NASA,in addition,the comparison experiment is done through comparing with software defect prediction models which were built by another four machine learning algorithms.The results show that the model proposed in this paper can reduce the false alarm rate and improve the prediction accuracy.Secondly,in order to further improve the classification accuracy of the software defect prediction model,the improved cuckoo search algorithm is proposed.After applying the improve cuckoo search(ICS)algorithm to the software defect prediction,the software defect prediction method based on ICS-ANN is proposed.The experiment results show that the proposed method further improve the classification accuracy of software defect prediction model,In summary,the proposed method can build software defect prediction model with high performance effectively,which will play a positive role to improve software quality and reduce software testing costs.