We use molecular simulations and machine learning to determine the vapor-liquid equilibria (VLE) properties of polycyclic aromatic hydrocarbons (PAHs). First, we perform Hybrid Monte Carlo Wang-Landau (HMC-WL) simulations on reference PAHs and evaluate VLE properties including the boiling and critical points, as well as the parameters capturing the dependence of vapor pressure upon temperature. Second, we use machine learning (ML) to extend the simulation predictions to a series of PAHs, with varying shape and number of fused rings. More specifically, HMC-WL results are used to train an artificial neural network using as descriptors molecular data, i.e. molecular weight, moments of inertia and number of aromatic rings. The combined molecular simulation-machine learning approach proposed here provides a self-consistent method to determine the VLE properties of organic compounds. (C) 2018 Elsevier B.V. All rights reserved.