Application Of Intelligent Algorithm In Price Analysis Of Traditional Chinese Medicine

As the concept of traditional Chinese medicine is gradually accepted by the world,the demand for traditional Chinese medicine is increasing year by year.Price analysis plays a very important role in the stability and development of the traditional Chinese medicine market.The price data of Chinese medicinal materials are affected by multiple factors such as growth cycle,demand cycle,seasonal change and policy,etc.It is a chaotic series with linear and nonlinear components,which is difficult to be predicted by conventional methods.The combination model based on ARIMA-LSTM is proposed to predict the price of traditional Chinese medicine,at the same time,correlation coefficient is used to analyze the correlation of it.The price of traditional Chinese medicine contains linear and nonlinear characteristics.This paper uses the combination model to divide the data into two steps for analysis and prediction.For linear features,ARIMA(Auto-Regressive Integrated Moving Average)model was used for prediction.The ARIMA model removes the nonlinear components irrelevant to the linear prediction by difference,and predicts the linear trend of the price in the future by the historical price of traditional Chinese medicine.LSTM(Long-Short Term Memory)neural network was used to predict the nonlinear characteristics.LSTM neural network has forgetting gate and memory unit,which is suitable for dealing with time series problems with nonlinear characteristics.In this paper,Monte Carlo method is used to improve the gradient descent method of LSTM to accelerate the convergence speed.LSTM neural network can better fit the nonlinear characteristic change trend of traditional Chinese medicine's price data.The idea of ARIMA-LSTM combined model is to make linear prediction with ARIMA first,subtract the linear prediction value of ARIMA from the original data,and then the remaining value contains the residual value of the nonlinear component.Then the LSTM neural network is used to predict the nonlinear residual value,and the prediction results of the two models are accumulated to obtain the prediction result of the combined model.Experimental results show that compared with single ARIMA and LSTM,the combined model has higher prediction accuracy and better stability.The innovation of this paper is to use correlation coefficient to analyze the correlation between different traditional Chinese medicines.In the past,the focus of price data prediction has always been around the price data itself.Traditional Chinese medicines are complementary and mutually exclusive in function,which make them highly correlated in price data.In this paper,the correlation coefficient method is proposed to improve the input of LSTM neural network,that is,the strength of correlation between medicinal materials is analyzed first,and other medicinal materials with high correlation are taken as its secondary reference factors,so as to constitute new input data.The experimental results show that compared with the unimproved combination model and support vector regression,the improved combination model using correlation coefficients has better prediction results.