Neural network analysis of spectroscopic data of lycopene and beta-carotene content in food samples compared to HPLC-UV-vis.

In this study a neural network (NN) model was designed to predict lycopene and beta-carotene concentrations in food samples, combined with a simple and fast technique, such as UV-vis spectroscopy. The measurement of the absorbance at 446 and 502 nm of different beta-carotene and lycopene standard mixtures was used to optimize a neural network based on a multilayer perceptron (MLP) (learning and verification process). Then, for validation purposes, the optimized NN has been applied to determine the concentration of both compounds in food samples (fresh tomato, tomato concentrate, tomato sauce, ketchup, tomato juice, watermelon, medlar, green pepper, and carrots), comparing the NN results with the known values of these compounds obtained by analytical techniques (UV-vis and HPLC). It was concluded that when the MLP-NN is used within the range studied, the optimized NN is able to estimate the beta-carotene and lycopene concentrations in food samples with an adequate accuracy, solving the UV-vis interference of beta-carotene and lycopene.