A hybrid model-based and data-driven framework for the screening of promising solvents is suggested to optimize reaction rates. The framework comprises a sequence of two connected problems: (i) identification of model to predict solvent effects on reaction rate constants from experimental data and (ii) computer-aided screening exploring a databank of solvents. The resulting problems are formulated as systems of linear equations which can be solved by standard numerical linear algebra packages. in light of the uncertainty inherently presents in experimental data, a combination of Tikhonov regularization and optimal design of experiments (or data selection) is proposed to remedy uncertainty amplification from the data to the solution and circumvent unreliable screening. The results obtained using the proposed strategy are compared with the benchmark solvent selection procedures. They are shown to be in good agreement with experimental data, in relative as well as in absolute terms, for the investigated case study, i.e., the solvolysis of tert-butyl chloride, which belongs to the class of S(N)1 reactions. (C) 2013 Elsevier Ltd. All rights reserved.