A Deep Reinforcement Learning Algorithm to Project Portfolio Selection Problem

Project construction and development are an important part of the development of any large organization. The most critical issue for decision-makers is how to select the right projects from the large number of candidate projects for construction and rational arrangement, that is, project portfolio selection and scheduling problem (PPSS). PPSP is a typical NP-hard problem. As the size of the problem grows, the effectiveness and practicality of traditional exact mathematical programming methods is greatly reduced due to the high time cost. Therefore, in this paper, we design a deep reinforcement learning method to solve this problem. We validate the algorithm on several datasets and compare it with metaheuristics for combinatorial optimization problems to verify the effectiveness of our algorithm.