PUGH Decision Trapezoidal Fuzzy and Gradient Reinforce Deep Learning for Large Scale Requirement Prioritization

Abstract

Objective: To prioritize requirements for large scale software projects within time involving uncertainty in the opinions among different stakeholders. Methods: We propose Pugh Trapezoidal Fuzzy and Gradient Reinforce Learning (PTF-GRL) methods for large scale software requirement prioritization. A Pugh Decision-based Trapezoidal Fuzzy Requirement Selection model is designed, inputting the functional and non-functional requirements of the corresponding stakeholders. With the assistance of Trapezoidal Fuzzy Inference, the qualitative factors are mapped with the corresponding numeric factors, which increases the computational efficiency. Findings: Performance is analyzed based on four parameters: The first parameter is accuracy and our method showed improvement of 4%, 7% and 3% compared to JRD-SCRUM, IFS and SRPTackle respectively. The second parameter is prioritization time and found that our method had reduced time of 30%, 37% and 39% compared with existing methods. The third parameter is precision and it was found that our method improves precision by 6%, 10% and 5% compared with the other two methods. The final parameter we consider is the test suite execution and our method showed improvement of 12%, 19% and 5% compared with the existing two methods. Novelty/Applications: The originality of this work indicates the better performance along with the optimal test suite execution even considering the uncertainty factor in the proposed method compared with existing similar methods. Keywords: Software Project; Pugh Decision Matrix; Trapezoidal Fuzzy Inference; Gradient Orientation; Reinforce Learning; Requirement Prioritization