Requirements Prioritization and Next-Release Problem under Non-additive Value Conditions

Abstract

Next Release Problem (NRP) is a complex combinatorial optimization problem consisting of identifying a subset of software requirements maximizing the business value under given constraints such as cost and resource limitation, time and functionality related dependencies between requirements. NRP can be mathematically formulated as an integer linear programming problem and previous researches solve the NRP multi-objective optimization problem using exact and metaheuristic search techniques. We present a mathematical formulation of the NRP under conditions of non-additive customer valuations (positive and negative synergies) across requirements. We present a model that allows customers to state their preferences or valuations across bundles or combinations of requirements. We analyze the economic efficiency gains, cognitive and computationally complexity of the proposed model. We conduct experiments to investigate the applicability of multi-objective evolutionary algorithms (MOEAs) in solving the NRP with non-additive valuations and implication constraints on requirements. We compare and contrast the performance of state-of-the-art MOEAs such as NSGA-II and GDE3 on synthetic dataset representing multiple problem characteristics and size and present the result of our empirical analysis.