A Hybrid Multi-Objective Evolutionary Optimization Algorithm for Next-Release Problem

Abstract

The software requirements selection process, a crucial step in software development, is a complex task. It involves identifying the most beneficial set of requirements for a software release while adhering to budget constraints. This problem, known as the next release problem (NRP), is challenging and classified as a non-deterministic polynomial (NP) hard problem. Interdependencies and other limitations further complicate the specified criteria. Selecting a specific set of requirements for the upcoming software release is a computationally challenging issue, falling under the category of NP-Hard problems. This paper proposes a hybrid method called MO-ACO-DE, which combines Multi-Objective Ant Colony Optimization with Differential Evolution to solve the multi-objective NRP. This work defines the NRP as a multi-objective optimization problem with two challenging objectives: customer satisfaction and development cost. Additionally, three constraints are introduced to address two real-world instances of the NRP. The proposed approach combines the management techniques of ant colony optimization (ACO) with the operators of the differential evolution (DE) algorithm to balance the exploitation and exploration stages of the optimization process. Both benchmark and real-world classic and realistic datasets were used for the experimental analysis of the proposed model. The results indicate that MO-ACO-DE outperforms other methods and enhances the fairness of requirement selection, especially when budget limitations are decreased.