Software Aging Analysis Based On Multiple Linear Regression Algorithm

Nowadays,we have seen a rapid development of science and technology.People become more and more dependent on computers and various computer applications in their daily lives.However,with the requirement for software performance increasing,software has become more and more complicated.Many problems inevitably occur after software continually run for a long time.Software aging is one of these problems.When software aging occurs after a long time run,system performance degrades,service quality declines,operation speed decreases,and even the whole sytem breaks down.Software aging is caused by the exhaustion of the resource,memory leak,unreleased file resources,storage space fragments,and internal error condition accumulation.If software can't supply steady,reliable and efficient services,it would cause unpredictable loss.Therefore,it is necessary to study problems of software aging.This paper uses the average load and response time as the criteria of software aging.The average load of the system can reflect the pressure of the current system,and the average response time can directly reflect the processing capacity of the current system.We judge software system aging by predicting the average load of the system,which provides a theoretical basis for the use of software regeneration strategy.First of all,this paper collects the important performance parameter data from the longtime running software systems.Suitable feature set is set up by means of feature selection.Then,we build the prediction model of software aging and predict the running state of the system based on the multiple linear regression algorithm.The experimental results show that the prediction model based on multiple linear regression can predict the changing trend of the system load,but the generalization ability needs to be further improved.In order to solve this problem,we further put forward a multiple linear regression model based on AdaBoost algorithm.This model taken the multiple linear regression algorithm as the base learner,and its accuracy of the prediction has been improved by integrating the learning characteristics of the AdaBoost algorithm.The experimental results show that the model has good generalization ability and it can improve the accuracy of prediction.