Prediction Of Financial High-Frequency Time Series Based On Improved Capsule Network

Financial high-frequency data is real-time data collected at a higher frequency and can reflect changes in the financial market.Stock trading data is a common high-frequency financial data.The short-term prediction of stock prices has always been a hot topic in the financial market,and accurately predicting stock prices can help investors better formulate corresponding investment strategies.As one of the important fields of artificial intelligence,deep learning has been widely applied in the financial market.In recent years,capsule networks,as a new type of deep neural network,have been widely applied in the field of computer vision and have achieved remarkable results.However,this model has not yet been widely applied in the financial field.Therefore,this article proposes to apply capsule networks to high-frequency financial prediction.The first part of this article uses a capsule network to conduct short-term prediction analysis of the closing price of Wuliangye at a five minute frequency.The experimental results show that the capsule network has good fitting ability for financial high-frequency data.To achieve more accurate fitting results,empirical mode decomposition is introduced to decompose and reconstruct the data,and then the capsule network is improved by removing the decoder and adding a fully connected layer as the output layer(the improved capsule network is labeled Eo Caps Net).The experimental results show that the EMDEo Caps Net model constructed in this paper effectively improves the prediction accuracy.The second part of this article uses Gram Differential Angular Field(GADF)to convert the five minute stock data of Wuliangye into image data,and uses capsule network to predict stock price fluctuations.To improve the training speed and stability of the training model,Eo Caps Net from the first part is used to classify and predict the GADF images of Wuliangye,and a convolutional neural network(CNN)is constructed as a comparative model.The experimental results indicate that the GADF Eo Caps Net model constructed in this article has a good predictive effect on the rise and fall of Wuliangye’s stock price,and the model is more stable.