Drimys angustifolia Miers is a tree species native to and found in southern Brazil. The extract of this plant is rich with active compounds that show medicinal potential, its uses being prospected as phytotherapy. In this study, yield data from supercritical extraction of D. angustifolia Miers are provided at different pressure and temperature conditions, and for various process operation times. Additionally, with the view to allowing a scale-up process, a methodology for identifying the extraction curves using neural networks based on wavelets was proposed. This showed good prediction performance provided that a sufficient number of extraction curves are used during training. The identification method proposed is robust, fast and optimal, in the sense that the best neural network structure and respective associated weights can be determined, thus optimizing a quadratic approximation criterion. (C) 2016 Elsevier B.V. All rights reserved.