Random subspace method for analysing coffee with electronic tongue

Electronic tongue (ET) systems based on nanostructured films and impedance spectroscopy have been successful in detecting trace amounts of impurities in water and distinguishing similar complex liquids, such as different wines and coffees. There is now the challenge to correlate the electrical response from an ET to taste as assessed by human experts. Employed is a hybrid regression approach based on the random subspace method and principal components analysis to predict the global quality of coffees. A high correlation between predicted scores and those assigned by expert tasters was obtained, which is represented by a Pearson coefficient of 0.964.