Exploratory Study Of Deep Learning Method In NMR Spectroscopy

With the rapid development of society,the demand for new drugs and new materials is becoming more and more urgent.However,in the analysis stage of organic synthesis mixture,it is extremely time-consuming and energy consuming,which slows down the synthesis process.On the other hand,in recent years,with the rise of artificial intelligence,especially the era of big data and deep learning,it has become the driving force behind the transformational development of many research fields under the background of interdisciplinary.This paper will explore the application of deep learning method in organ-ic synthesis.In the real task of organic synthesis,NMR is an important tool for structure identification.If we can use deep learning method to identify the structure of compounds based on the original NMR data,it will be expected to provide an important method for judging the mixture of organic synthesis.This will be a pioneering achievement and will greatly enhance the process of organic synthesis.Since there is no relevant research in this field,this paper will start from the data collection and sorting,and study the compound structure recogni-tion of NMR spectrum based on deep learning method.Firstly,this paper gives a detailed overview of the methods and related theories.Secondly,this paper ana-lyzes the different methods of feature collection of NMR data,and builds models on different data sets by using SVM and KNN.Then this paper analyzes and compares the results.After that,aiming at the research of data imbalance,this paper proposes an oversampling method based on the characteristics of nuclear magnetic data,and explains the deep theoretical basis of deep learning method.Eventually,in this paper,the deep learning model NMRClass based on recurrent neural network is innovatively constructed for NMR data.And the recognition of compound structure from the original NMR data is realized.The results show that the deep learning method NMRClass has the performance of low loss and high precision in the structure recognition of NMR spectra.In the future,it will be applied to deeper research to provide help for the structural identification of synthetic mixtures.