REIMAGINING MAINTAINABILITY MACHINE LEARNING TECHNIQUES FOR SECURITY REQUIREMENT OPTIMIZATION

Abstract

Software maintainability has gained recent popularity in the sphere of software engineering throughout the past few years in an effort to determine the quality of software. Hence, it is important to predict this maintainability in time and with accuracy for the effective administration of software during the maintenance stage. In turn, it is causing the developer to focus more on those modules that are expensive to maintain. software maintainability prediction (SMP) machine learning model suggested in this paper is informed by the Students project requirements software requirements dataset. This study is a description of the use of machine learning high-end methods of classification namely the Random Forest, AdaBoost and Voting Classifier which significantly contribute to the evaluation of maintainability in terms of software security requirements. To make a comparative analysis, these models have significant accuracy improvement with the highest accuracy of 87.85 in binary classification and 99.37 in multi-class classification compared to 79.43 and 85.08 of the baseline models respectively. The results show that more ML methods can be used to enhance the efforts to measure the maintainability of the software and that these methods can be used to meet the requirements of software security