Classification And Detection Of Raman Spectral Data Based On Machine Learning Algorithm

In view of the advantages of non-destructive,rapid,accurate and wide detection range,the detection method of food-borne pathogens has been gradually transformed into a new method of detection by Raman spectroscopy.Raman spectrum is a kind of scattering spectrum based on Raman effect,which can provide rapid,simple and non-damaging qualitative and quantitative analysis.Through the analysis of Raman spectrum,the vibration rotation of substances can be known,so as to identify and analyze substances,reduce errors and sample preparation time,and obtain higher sensitivity.In this paper,several samples of foodborne pathogens were collected,which were provided by Changchun Veterinary Research Institute,Academy of Military Science of the Chinese People’s Liberation Army.In this study,E.coli and Brucella were mainly used as experimental objects for Raman spectral data collection,denoising,smoothing,normalization and other pretreatment operations.The classification model is constructed through principal component analysis and linear discriminant analysis combined with machine learning algorithms such as gradient lifting tree and convolutional neural network,and different data are used for comparison experiments,aiming at reducing errors caused by sample feature number and providing a qualitative reference for classification detection and recognition by combining Raman spectrum with machine learning algorithm.The main content of this paper is as follows:The data of Raman spectrum were collected,the noise existing in Raman spectrum was removed by denoising smoothing method,and the data was further processed by standardization and normalization.Different dimensionality reduction methods--principal component analysis(PCA)and linear discriminant analysis(LDA)were applied to Raman spectral data.Different dimensionality reduction methods combined with machine learning algorithm classification model were used for comparative experimental analysis.The experimental results showed that the classification model could successfully classify the two foodborne pathogens and improve the classification accuracy on the basis of the original algorithm.A convolutional neural network classification model is proposed,which uses data enhancement to expand sample data,builds a convolutional neural network classification model for training,and compares and analyzes the original data and data after pre-processing and dimensionality reduction.Experiments show that the model can successfully classify sample data and has strong robustness.