Strategic Workforce Planning with Deep Reinforcement Learning

This paper presents a simulation-optimization approach to strategic workforce planning based on deep reinforcement learning. A domain expert expresses the organization's high-level, strategic workforce goals over the workforce composition. A policy that optimizes these goals is then learned in a simulation-optimization loop. Any suitable simulator can be used, and we describe how a simulator can be derived from historical data. The optimizer is driven by deep reinforcement learning and directly optimizes for the high-level strategic goals as a result. We compare the proposed approach with a linear programming-based approach on two types of workforce goals. The first type of goal, consisting of a target workforce, is relatively easy to optimize for but hard to specify in practice and is called operational in this work. The second, strategic, type of goal is a possibly non-linear combination of high-level workforce metrics. These goals can easily be specified by domain experts but may be hard to optimize for with existing approaches. The proposed approach performs significantly better on the strategic goal while performing comparably on the operational goal for both a synthetic and a real-world organization. Our novel approach based on deep reinforcement learning and simulation-optimization has a large potential for impact in the workforce planning domain. It directly optimizes for an organization's workforce goals that may be non-linear in the workforce composition and composed of arbitrary workforce composition metrics.