Software Defect Prediction Model Based On Artificial Fish Swarm Optimization Depth Sparse LSSVM

With the continuous extension of information technology in all walks of life,the scale and complexity of program software are becoming more and more complex.More comprehensive architecture design,a lot of code writing makes the subsequent functional testing,operation and maintenance work no longer easy.The software defect assessment and prediction is to verify the deficiencies and errors in the program system in a certain way,so as to lay a foundation for the staff to carry out functional testing and provide users with more convenient and smooth use of the system.Among the existing software defect prediction and evaluation methods,the Least Squares Support Veotor Machine(LSSVM)is used to study the measurement metadata and defect data.It has been proved that the software defect evaluation model constructed by this method is effective.Based on this research,the following research is carried out in this paper.In this paper,an Improved adaptive artificial fish school algorithm(AAFSA)was proposed to improve the accuracy of artificial fish school algorithm by studying the adaptive parameter change mechanism.The algorithm mainly updates the Visual,Step and parameters of the artificial fish swarm algorithm in a certain range adaptively,which reduces the time complexity of the algorithm and improves the efficiency of the algorithm search.In this paper,and depth for sparse LSSVM parameter selection difficulties in classification and common dimension reduction algorithm can lead to premature beneficial properties information screen in addition to the problem,by using artificial fish algorithm better global search ability and potential parallelism,put forward a kind of artificial fish optimization based on adaptive sparse LSSVM software defect prediction model(AAFSA-DLSSVM).In this model,the depth sparse LSSVM mainly plays the role of a classifier,and USES the improved artificial fish swarm algorithm to determine the optimal measurement attributes and the optimal parameters of the depth sparse LSSVM.Compared with the traditional research method,this model can retain the effective software defect measurement attribute data and lay a foundation for better enhancing the accuracy of software defect prediction.In order to prove whether the method proposed in this paper is effective,on the one hand,the paper compares AAFSA with other algorithms to optimize the benchmark function,and fully demonstrates its advantages.In addition,the software defect data set published by NASA was used for specific simulation tests,and aafsa-dlssvm was compared and analyzed with other traditional methods.The experimental results show that the AAFSA algorithm has better optimization effect and convergence speed than other algorithms.The DLSSAVM model based on the optimization of AAFSA has a better effect on various performance evaluation indexes of software defect prediction,which is more prominent than the prediction model.