Selecting balanced portfolios of R&D projects with interdependencies: A Cross-Entropy based methodology

•We propose a 0–1 nonlinear integer programming model to balance values and risks in R&D project portfolio selection.•This model considers objective of maximizing portfolio value while considering various types of R&D projects, their risks and interdependencies.•In order to solve the proposed model, we present a Cross-Entropy algorithm for a R&D project portfolio selection problem.•The algorithm proved to be very effective, in terms of solution quality as well as computational time.

The intensive growth of technology makes firms rely on research and development (R&D) activities in order to adapt to technology changes in an ever-changing and uncertain environment. Due to R&D budget constraints and limited resources, firms are often forced to select a subset of all candidate projects by means of project portfolio selection techniques mitigating the corresponding risks and enhancing the overall value of portfolio. Projects' interdependencies and types were rarely considered in existing models of R&D portfolio selection that may result in selecting wrong projects. This flaw hinders the projects alignment with corporate objectives and strategy and leads to excessive risk and missing the promised values. In this paper, a balanced set of R&D project evaluation criteria was proposed. Next, to construct R&D project portfolio, a 0–1 nonlinear mathematical programming method for balancing portfolio values and risks was proposed, in which research projects' interdependencies, types and other constraints were all considered. Finally, a Cross-Entropy algorithm was developed to solve the proposed model and results were reported. The algorithm proved to be very effective in terms of solution quality and computational time. The proposed algorithm especially suits large scale instances while exact approaches are doomed to fail.