Drug Classification and Identification Method Based on X-ray Absorption Spectroscopy Combined with Machine Learning

Abstract

With the emergence of new drugs, it is imperative to study new detection methods with higher detection speed and accuracy. Traditional detection methods require high requirements for detection instruments and environments, and are complex to operate. x-ray absorption spectroscopy NDT technology has the advantages of low operational difficulty, penetrating observation, and strong ability to differentiate substances, and is well suited for drug detection and identification. A combination of X-ray absorption spectroscopy, convolutional neural network (CNN), support vector machine (SVM) and improved particle swarm optimization algorithm (IPSO) is used to achieve the classification and identification of drugs. Firstly, 14 chemical reagents with chemical formula similar to that of drugs are selected as experimental samples, and the X-ray absorption spectra of these 14 samples are obtained using X-ray detectors. Then, the features of the spectral data are extracted using CNN, and the SVM model is trained with the extracted features and the two important initial parameters of the SVM are optimized using IPSO with the introduction of weight decay. Finally, the trained model is applied to the test set and the performance is evaluated by several metrics. The experimental results show that the model not only makes the parameter-optimized traditional machine learning model SVM very effective and improves the prediction accuracy to 99.14%, but also avoids the disadvantages of high complexity, dramatically long running time and reduced efficiency due to the direct fusion of IPSO and SVM, and can be well applied to drug classification, which can substantially improve the recognition accuracy with almost constant The algorithm running efficiency can be improved substantially with almost the same recognition accuracy. Therefore, the combination of X-ray absorption spectrometry with CNN, IPSO and SVM can provide a fast, highly accurate and reliable classification and recognition method with broad application prospects in the field of drug detection and identification. CCTS: O611.5 Literature Identification Code: A