Identification of syrup adulteration in wolfberry honey using CNN-CBAM-SVM combined with ¹H NMR

To identify syrup adulteration in honey, a deep learning model based on the CNN-CBAM-SVM architecture combined with ¹H NMR spectra was developed. The traditional CNN model was enhanced by incorporating the CBAM module and replacing the fully connected layer with an SVM classifier, making it well-suited for small sample sizes. The ¹H NMR spectra of 20 genuine and 20 adulterated samples with two distinct syrups were divided into a training set (32 samples) and a validation set (8 samples). The CNN-CBAM-SVM architecture effectively addressed issues such as category imbalance in the softmax layer, weak generalization, poor robustness, and low detection accuracy typically encountered in traditional CNN models. The final model achieved 100 % accuracy on both the training and validation sets, and t-SNE visualization further confirmed the model's correct classification performance. Therefore, the integration of ¹H NMR spectra with the CNN-CBAM-SVM model holds significant potential for detecting syrup adulteration in honey.