Synergistic pore size screening and dual time window kinetics: machine learning-based discrimination and concentration prediction of different types of preservatives in complex food matrices.

Distinguishing between different preservatives is of great significance in quality control. Therefore, developing effective methods to distinguish between them is of great importance. This study overcomes the limitation of traditional sensor arrays that identify only a single analyte. Through the synergistic integration of pore size screening, metal doping, and machine learning, this system achieves multi-category discrimination of different preservatives, with a minimum discrimination concentration of 150 nM. With the optimal pore size effect of the ZIF-8 nanozyme and a dual-window reaction analysis strategy, its anti-interference capability and discrimination performance are significantly enhanced. Adsorption kinetics and electron paramagnetic resonance experiments reveal the mechanism underlying the discrimination between different preservatives. Furthermore, by integrating machine learning models, the system enables both concentration-independent identification and concentration regression through model switching. The proposed detection model exhibits good discrimination performance in real sample matrices.