Research On Aging Prediction Of Software System Based On Least Squares Support Vector Machine

With the long-term continuous operation of the software system,the defects left in the software will cause the phenomena of computer memory leakage,accumulation of rounding errors,and file lock failures to be released in time,eventually leading to deterioration and even collapse of the system performance.The occurrence of these software aging not only reduces the reliability of the system,but also endangers the safety of human life and property.In order to reduce the damage caused by software aging,it is particularly important to predict the trend of software aging and adopt an anti-aging strategy to avoid software aging.A very important indicator that reflects the state of software aging is the average software response time.This dissertation infers software running state by predicting the software average response time.Aiming at the research of software aging prediction,the main work in this dissertation is summarized as follows:Based on the analysis of the existing prediction methods and the operation characteristics of the software,it is determined that the least squares support vector machine is used as the prediction method of software aging.Experiments are carried out to build the software aging prediction model of least squares support vector machine.The data from the experiments are compared with those from the software aging prediction model of support vector machine,and the superiority of the former model is proved.The parameter-selections of penalty factor and kernel function in least squares support vector machine prediction model have a great influence on the prediction effect of the model.In order to improve the prediction result of the least squares supportvector machine,this dissertation adopts the gray wolf optimizer to tune the parameters of the least squares support vector machine.The experiment results are compared with those from the model whose parameters tuned by genetic algorithm to prove the former's superiority.The solution can provide an important reference for determining the timing of execution of software anti-aging operations.