Impedimetric Hydrogel Sensor for the Identification of Hexose Using Machine Learning

Determination of the type of sugar present in a beverage is one of the key parameter in assessing their nutritious values. This article presents a novel technique of hexose identification using hydrogel sensor in a sugar sample. The hydrogel uses poly(vinyl alcohol) (PVA) and benzene-1,4-diboronic acid (1,4 BDBA) whose boronate ester bonds break down when they come in contact with sugar samples and change the gel impedance. A tree-construct algorithm is proposed to identify the hexose type, irrespective of its concentration from the impedance data of the proposed gel at 20 and 200 kHz. The algorithm includes 22 machine learning (ML) models for 22 different scenarios. The ML models are designed using MATLAB ML classifier toolbox based on 7452 datasets and 414 samples. The proposed algorithm is further validated using 300 test samples, and it is found that the true positive rate (TPR) in identifying fructose, glucose, galactose, and sucrose is 84.3%, 66.3%, 69.7%, and 80.7%, respectively. This indicates that the proposed hydrogel is more reliable for fructose and sucrose identification.