Starting with historical demand trajectories of similar products and their covariate information, we develop a data-driven approach to dynamic inventory planning for new products. Specifically, we employ non-parametric machine learning methods to model the uncertain future demand and incorporate...