Document Type : Original Article

Author

• Keyvan Asefpour Vakilian

Assistant Professor, Department of Biosystems Engineering,, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran.

Abstract

During the last two decades, various types of enzymatic biosensors have been introduced for the specific and selective detection of nitrate. These biosensors generally use the redox reaction of nitrate-nitrite to measure nitrate. Since the activity of the enzyme used in the biosensor structure decreases over time, the user of the biosensor should frequently replace the enzyme immobilized onto the working electrode, which increases the detection costs and limits their commercialization. In this study, artificial neural networks (ANNs) have been used to predict nitrate concentration in samples by considering electrochemical data and the decrease in enzyme activity over time. The Harris hawks algorithm was used as a meta-heuristic optimization method to optimize weight and bias hyperparameters of ANNs used in the biosensor decision-making unit. The results showed that the optimized learning algorithm led to a promising prediction of nitrate concentration at the micromolar level with a coefficient of determination of 0.95. In addition, the introduced biosensor could be used up to 30 days after enzyme immobilization. A comparison between the findings of this study and previous studies that used support vector machines and fuzzy inference systems showed that ANNs optimized with novel meta-heuristic techniques can provide more reliable prediction results.

Keywords

• Collective intelligence
• electrochemical data
• enzyme
• Harris hawks algorithm