A Machine Learning Algorithm Based on Inverse Problems for Software Requirements Selection

Abstract

This paper describes the software requirements prioritization task and provides a systematic approach to determine what needs to be included in the next release of a software product. Minimizing the total cost of adding a new feature in the next release and maximizing overall customer satisfaction are contradictory objectives. In this paper, first, an adaptive multi-objective prioritization model is discussed. Then we describe how discrete inverse problems ideas can in fact be formulated to obtain a smooth local “Added Degree of Importance” (ADI) function of client requirements which could be used to classify and prioritize the software requirements for next release. The numerical implementation of the proposed model with a case study on software requirements selection shows the effectiveness of the multi-objective inverse model (IM) approach. The proposed model have been compared with some of the recent relevant models. Main future of the model is that it has been designed by the assignment of a real score for each of the requirements unlike just classification provided in the literature.