Four machine learning models are used for making the prediction model while training is performed simultaneously using static as well as dynamic metric suite. Using prevalent prediction results are analyzed. For practioniors this result will be helpful. In this paper for maintainability prediction there is analyses of the effectiveness of dynamic metrics.
Open source was used to measure the capability of dynamic metrics. To develop prediction model four machine learning algorithms are used which are Linear Regression, Multilayer Perceptron, Gaussian Process and SMOreg. There are two independent variables dynamic metrics suite and another with static metrics suite.
In this paper the results of prediction using static metrics and dynamic metric deployed with all machine learning algorithms using prevalent prediction accuracy measures such as mean absolute error (MAE) and root mean square error (RMSE) are used . Linear regression was most accurate in learning models among all the four machine learning. We conclude that for maintainability predictions. dynamic metrics can be used. A small set of software code is linked to reliability .For the understandability a plug in is created which calculates and aggregates by it which produces a high-level interactive html report.
To solve the issue of software product quality a first step Plug-in was developed. Other step is to described plug-in as part of a research effort. Software maintenance includes modification of out of date abilities. Software change is required due to the adjustment in the essentials of the client, modification in the technology, change at the stage where the software will be deployed .
Change is unavoidable so mechanisms must be created for valuation, and taking charge of modifications. Maintenance is to change the code after the change. After the delivery of software.The design is to protection the estimation of software .Effort of generating the software is less then the effort of keeping up the software. Maintaining software is hard than creating the software.
At the development stage cost can be controlled by checking software metrics. Different points of software such as polymorphism , abstraction are the quantitative estimation. At every phase of sdlc software metrices are calculated. Static metrics just catch the physical conduct of the software, genuine code isn’t accomplished. By the use of dynamic inquiry metrics are estimated. With the support of dynamic metrics, we can determine answers to a large number of question like, what number of lines of code are getting executed, which methods have been called at run time and which class is coupled to other class, and so on.
In this article, dynamic metrics have been used for the estimation of software maintainability.