Abstract
Rapid inspection of adulterated dietary supplements containing appetite suppressants seized at the scene can provide clues and directions for case investigations. Meanwhile, applying filters and feature extraction techniques can effectively remove noise and interference from the original spectral data, thereby improving the recognition performance of the model. In this study, a total of 419 samples of 7 appetite suppressants confiscated from actual cases were collected. Fourier and Hilbert transform filters were used to preprocess the original spectral data of the samples. Feature extraction was performed using competitive adaptive reweighted sampling, and classification models were established using partial least squares discriminant analysis and random forest algorithms for identification purposes, aiming to select the optimal filter for noise removal. The results showed that the recognition rate and stability of the original spectral data significantly improved after filter processing, with the Hilbert transform filter demonstrating the best noise removal effect. The random forest model outperformed the partial least squares discriminant analysis model regarding recognition rate and stability. The optimized model achieved a recognition rate of 99.21%. This method effectively filters out noise from the spectral data using filters, enhancing the qualitative identification capability of the model and providing reference value for the rapid identification of appetite suppressants in forensic science.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call