The attrition rate in the development of new drugs is exceptionally high with poor early prediction of membrane permeability being a contributing factor. For this reason, experiments aimed at early identification of drug candidates with poor membrane permeability are commonly used in drug development. A novel method for testing permeability levels and generating analyzable data is through the usage of microcavity array supported lipid bilayers connected to an electrochemical cell to generate output. However, this approach is time consuming meaning that relatively small datasets are generated from the process. The goal of this research is to optimize data generation by narrowing the frequency range of the Working Electrode part of the electrochemical cell and to do so using whitebox machine learning functions. A robust validation is carried out which classifies a total of nine drugs, identifying a range of permeability strengths at very high accuracy levels. The misclassification of drug types is fully explainable through the provision of simple rules for each incorrect classification.