Multichannel electrochemical workstation-based data collection combined with machine learning for online analysis of tyrosine

Abstract

The low reliability and accuracy of electrochemical workstations significantly restrict their wide application in industrial and residential inspections. In this study, we combined the machine learning and multichannel detection methods to filter and model massive amounts of data collected from active electrodes to extract usable data and alleviate the aforementioned disadvantages of electrochemical workstations. We constructed carbon–graphene oxide (CB–GO) composite electrodes on screen-printed substrates for tyrosine (Tyr) monitoring. Subsequently, an electrode matrix comprising multiple identical electrodes (CB–GO/SPCE) was replicated and mounted on a designed piece of multichannel electrochemical workstation. We used an on-board program to design various electrochemical methods for the electrochemical detection of Tyr. We used the CB–GO/SPCE electrode array for testing the Tyr differential pulse voltammetry data. The collected data were first subjected to human screening. Next, a combination of t-test and false discovery rate methods was used to further process the data and reduce their dimensionality before using them as the input for the constructed model. We selected K-fold validation with cross-validation data for the input data owing to the relatively small amount of available data. We used an electrochemical testbed with CB–GO/CP electrodes for online analysis of Tyr. Subsequently, a small artificial neural network was constructed to analyze the data. The analytical results were plotted against theoretical targets to examine the concentration of the substance with corresponding loss and accuracy curves. The results revealed that the neural network model reasonably predicted the Tyr concentration on a completely new validation data set based on the data collected from the designed electrochemical test platform. The proposed multichannel portable electrochemical analysis platform combined with machine learning can be widely applied to electrochemical analysis methods to improve their reliability.