Study On Spectrum Of Chinese Medicine Based On Functional Data Analysis

In this paper,we study the UV spectrum of Flos Lonicerae Japonicae solution,which is one of the raw material of Qingkailing injection.Firstly,we build a functional linear regression model between its concentration and the UV spectrum data to explore a rapid quality detection method.Secondly,considering the problem that the sample size are too small in the empirical study,we propose two methods to increase the sample size.One is the sample-split method and the other is the Bootstrap method.Thirdly,we use Monte Carlo experiment to test how their results performed and make a comparison.Comparing the results of the Monte Carlo experiment,we find that when the functional independent variable measured densely,the sample-split method cannot improve the estimation precision of the model parameters.Besides,whether the measurement-error estimation method is used has a very limited effect on that.However,the Bootstrap method can improve the estimation to some degree.At last,applying the above analyze method to the empirical part,we built a functional linear regression model for the concentration of Flos Lonicerae Japonicae solution on its UV spectrum.Out-of-sample forecast is made in the paper to compare with the existing method.Based on the results of this paper,we reach the conclusion that applying functional data analysis method to the UV spectrum data of Flos Lonicerae Japonicae solution can use the information contained in data itself to the greatest extent,and analyze the functional data from an overall and dynamic view.Besides,by comparing the concentration prediction method,which is building the functional regression model between the UV spectrum and the concentration of solution,with the existing methods,we also find that the prediction accuracy of the former method has improved significantly.Moreover,when the sample size is small,Bootstrap method can help to solve this problem.