Tree-based machine learning models assisted fluorescent sensor array for detection of metal ions based on silver nanocluster probe

[Display omitted] •The sensor array is constructed by only a single sliver nanoclusters probe.•Fluorescence spectra were analyzed by multiple tree-based machine learning models.•LDA was used to reduce the dimensionality of data and improve model performance.•Six metal ions and their binary mixtures are successfully identified.•The concentration of each metal ions was precisely predicted by tree-based models. The growing concern over heavy metal pollution and its impact on the environment and human health has led to a proliferation of research on the detection and differentiation of heavy metal ions. A novel fluorescent sensor array utilizing only one single Ag-nanoclusters (Ag NCs) was developed for the efficient detection of six metal ions. The Ag NCs probe was prepared by using poly(methyl vinyl ether-alt-maleic acid) (PMVEM) as the ligand and has different fluorescence properties in water and dimethyl sulfoxide (DMSO). The interaction between metal ions and Ag NCs resulted in a characteristic fluorescence variation pattern which was subsequently analyzed using various tree-based machine learning models. We have compared different combinations of classification models and pre-processing methods of which the K-Nearest Neighbors Classifier with the first five linear discriminants has the highest accuracy. Through the integration of concentration models within a tree-based pipeline optimization framework, six unique concentration regression models were selected for each metal ion. In addition, the developed sensor array could identify metal ions in binary mixtures. And it still kept high accuracy for the classification of six target metal ions in river water. In conclusion, the proposed framework was found to be effective in the detection of heavy metal ions in environmental samples, thus providing a promising approach for addressing heavy metal pollution.