The recent trends in the ICT industry are related to the development and implementation of green software products and practices. The way to seek for energy efcient solutions starts from the proper and precise assessment of the current state of the system in hand. In this paper we present the software-based approach to the energy efciency assessment. We propose machine learning based approach to estimate energy consumed by a computer system. We evaluated our approach on datasets extracted from systems running on Linux and Windows operating system. Using MSE, MAE and R 2 our energy consumption estimation model reached 0.0007, 0.0104, 0.9214, respectively.