Research On Forecast Of Sanchi Price Trend Based On Deep Learning

The dramatic price fluctuations of Sanchi severely affected the healthy development of the Sanchi industry.Accurately predicting the trends of Sanchi price fluctuations can provide a valuable reference for the majority of Sanchi practitioners.Based on the price time series of relevant Chinese herbal medicines in the compatibility law of Sanchi,this paper uses the deep learning method to predict the trend of Sanchi's price.The main contents of this article are as follows:First,a Long Short-Term Memory is used to predict the price fluctuation trend of Sanchi.The price fluctuation of Sanchi is not only affected by the historical price of Sanchi,but also by the historical price of related traditional Chinese medicines.Therefore,based on the price time series of 29 kinds of traditional Chinese medicines in the compatibility law of Sanchi,LSTM is used to learn the effect of price changes of related traditional Chinese medicines on the price of Sanchi.The slope time series is used to represent the price fluctuation trend,and the Dynamic Time Warping(DTW)is used to measure the degree of similarity between the price fluctuation trend of 29 Chinese medicinal herbs and the price fluctuation trend of Sanchi.Then,by setting similarity thresholds,price fluctuation sequences with different similarities are selected as input features of LSTM in related Chinese herbal medicines,and the prediction accuracy of LSTM under different input feature sets is evaluated,so as to select the best input features of the prediction model and predict the price fluctuation trend of Sanchi.Experiments show that the price trend prediction method of Sanchi based on LSTM considers the influence of related traditional Chinese medicine on the price of Sanchi,and its prediction performance is obviously better than the traditional price prediction method.Second,a hybrid neural network is used to predict the trend of Sanchi's price.The historical price fluctuation sequence determines the direction of the price change of Sanchi.Meanwhile,the future price of Sanchi also changes on the basis of the current price.Therefore,this paper predicts the trend of Sanchi' price by constructing a hybrid neural network.On the one hand,the hybrid neural network uses the LSTM module to learn the long-term dependence of the price fluctuation trend of Sanchi,on the other hand,it uses the CNN module to learn the local characteristics of the original price corresponding to the trend.Then,the hybrid neural network uses a feature fusion layer to map the characteristics of the two into the same space for fusion,and finally outputs the prediction through a full connection layer.The experimental results show that the hybrid neural network has better prediction performance than the single LSTM and CNN models because it integrates the long-term trend and short-term characteristics of the price.