Raman Spectrum Recognition Based On Machine Learning Algorithm

With the continuous improvement of human living standards,some dietary structures and environmental conditions of people have changed,resulting in a gradual increase in the incidence rate of some diseases in the population.For example,diabetes has become one of the common chronic diseases in the world.diabetes patients need long-term disease management.If they are not found and treated in time,they will face the threat of many complications,causing extremely serious physical and mental distress,giving individuals Families and society bring heavy burdens.Therefore,timely screening and diagnosis of certain diseases have a crucial impact on human health.At present,multiple experimental studies at home and abroad have shown that Raman spectroscopy technology has been widely used in biomedical detection,such as the detection of biological fluid samples(urine,blood,cerebrospinal fluid),cancer biomarkers,and cell tissues.This technology has the advantages of non-invasive,anti-interference,and relatively short collection time,regardless of whether the sample is solid,liquid,gas,ointment,or powder,Raman spectroscopy can quickly characterize its chemical composition and structure.Therefore,in this paper,Raman spectroscopy technology combined with machine learning algorithm is applied to the detection of novel coronavirus and screening of diabetes patients,giving full play to the advantages of Raman spectroscopy technology in speed,nondestructive,high sensitivity and high accuracy.The main work of this article is as follows:Study the principle of Raman spectroscopy generation,introduce the process of data collection in Raman spectroscopy experiments,and the spectral noise generated during the collection process.Use the smoothing algorithm Savitzky Golay method to smooth and filter the noise.During the experiment,continuously adjust the window width(Point of Window)to achieve the best effect of Savitzky Golay filter on spectral noise processing.Due to the significant difference between the Raman shift of the sample spectral data and the corresponding Raman intensity,this article chooses to use the Min max normalization method to perform linear transformation on the original data;Due to the large number of eigenvalues in the initially collected spectral data and the generally high sampling frequency,there is a high degree of coupling between each feature point.This article uses principal component analysis and linear discriminant analysis methods to reduce the dimensionality of the spectral data,reducing the dimensionality of the dataset and losing the main information of the data.In order to identify whether people are infected with COVID-19 and screen diabetes patients,this paper combines Raman spectroscopy technology with machine learning and deep learning algorithms,such as xgboost,convolutional neural network,K nearest neighbor,logical regression,support vector machine,etc.to build spectral classification recognition models,compare the recognition accuracy of different models,constantly optimize algorithms to improve the training speed of models,and detect COVID-19 Screening for patients with diabetes provides a non-destructive auxiliary method.