Machine learning assisted dual-emission fluorescence/colorimetric sensor array detection of multiple antibiotics under stepwise prediction strategy

Effective sensors to detect antibiotics as environmental and health hazards are urgently needed. Herein, we constructed a dual-emission fluorescence/colorimetric sensor array based on novel high fluorescence quantum yield carbon dots and CdTe quantum dots. Multi-dimensional data (fluorescence intensities and maximum emission wavelengths) was used to establish a sensor array platform with improved specificity. To meet the challenges of establishing a unified model and detecting outside datasets samples, we innovatively built a unified SX-model using a “stepwise prediction” strategy combined with machine learning to screen optimal methods. By integrating classification and concentration models under a tree-based pipeline optimization technique framework, the extreme random forest was selected as the most accurate classification model. The sensor array detected nine antibiotics at 0.5–50 μM with 95% accuracy and 4.93% average concentration error for unknown samples outside the datasets. Simultaneous identification of binary and ternary mixed samples was also enhanced. Furthermore, antibiotics in 216 river water and milk samples were discriminated with 100% accuracy and 3.25% and 4.43% average concentration errors of unknown samples outside the datasets, respectively. Finally, antibiotics were completed visually identified. The proposed original SX-model assisted dual-emission sensor not only overcomes low specificity and immobility, but breaks the bottleneck of existing analysis methods showing great application potential in the array detection field.