Molecularly imprinted electrochemical sensor arrays combined with machine learning for simultaneous determination of three neonicotinoid insecticides

Exposure to different neonicotinoid insecticides (NNIs) can cause varying degrees of harm to mammals and may even be carcinogenic. Due to their similar molecular structures, it is not only difficult to distinguish NNIs in analysis, but also cross-reactions can also occur. These cross-reactions cause the calibration curves to exhibit strong nonlinearities that cannot be fitted by usual mathematical models. Here, we present an electrochemical sensor array comprising three sensing units for the simultaneous determination of imidacloprid, thiamethoxam, and nitenpyram. The method eliminates cross-reaction with the aid of machine learning. The machine learning model comprises three components: the Douglas-Peucker algorithm for data compression, principal component analysis for classification, and an artificial neural network for quantification. The randomly assigned validation set showed a classification accuracy of 96.3 % for the model. The prediction accuracy was 98.77 %. The limit of detection was <0.037 µmol/L, with a detection range from 0.1 µmol/L to 200 µmol/L. Finally, the spiked tea samples were tested, and a satisfactory agreement was obtained between the expected and predicted values.