A Nitrate Enzymatic Biosensor based on Optimized Machine Learning Techniques

Many enzymatic biosensing devices have been developed for specific and selective nitrate detection over the last two decades. They generally use the nitrate-nitrite redox reaction to measure the nitrate. Since the activity of the enzyme used in the structure of the biosensor reduces over time, the operator should replace the enzyme immobilized on the working electrode frequently, which increases the detection costs and limits their commercialization capabilities. In this study, artificial neural networks (ANNs) have been used for nitrate concentration prediction in the samples considering both electrochemical data and enzyme activity decrement over time. Genetic algorithm (GA) and particle swarm optimization (PSO) were used to optimize the architecture of the ANNs used for the decision-making of the biosensor. This study showed that both GA-ANN and PSO-ANN learning algorithms resulted in promising nitrate prediction performance with coefficients of determination (R
²
) higher than 0.93 and 0.94, respectively. Moreover, the biosensor could be used up to two weeks after the enzyme immobilization for nitrate determination in more than 500 samples after each time of electrode preparation. Finally, a comparison between the findings of this study and former studies that used support vector machines and fuzzy inference systems revealed that ANNs optimized with evolutionary and swarm intelligence techniques can provide more reliable prediction results.