Research On Stock Price Prediction Based On Singular Spectrum Analysis And Temporal Convolutional Network

The stock market is an important part of the world's financial industry.At the same time,as a barometer of the country's macroeconomic operation,the price changes in the stock market also directly affect the stability of the financial market and the sustainable development of the economy.In the stock market,the stock price is an important reference indicator for analyzing the future development trend of the stock market,and the prediction of the future trend of the stock price has also received extensive attention,so the research direction of this topic has certain practical significance.With the rise of machine learning and deep learning in recent years,neural network models for stock price prediction have been continuously developed,which has improved the limitations and drawbacks of traditional prediction models.In previous studies,many scholars have proposed a variety of models for the problem of stock price prediction.There are traditional prediction methods based on statistics and methods based on neural networks.Although traditional models have high stability,they are not good at predicting high-frequency data such as stock prices.At the same time,it is difficult to process data with many dimensions and large quantities.Although a single neural network model works well when processing large-scale data,its stability is average.In many studies,many scholars have used recurrent neural networks,long and short-term memory networks and their related variant networks to predict stock prices.Although the effect is good,such networks are time-consuming and unstable during the training process.At the same time,there is also the problem of over-fitting.In the latest research,it is found that a convolutional neural network with a specific structure can predict the sequence well and improve the shortcomings of other networks in predicting stock prices.Therefore,this paper proposes a combined prediction model based on singular spectrum analysis and time convolutional network.Use relatively traditional data denoising methods to process the original stock data,retain long-term trend characteristics,reduce data jitter in a short period of time,and use a special convolutional neural network to predict the sequence.In the neural network prediction experiment,we design two types of experiments.On the one hand,we use short-term grouping experiments to verify the accuracy of the time series convolutional network for stock price prediction,and on the other hand,we use long-term prediction experiments to verify the trend of the time series volume network for stock price changes.The accuracy of the prediction,and finally through the analysis of the experimental results,we found that the new model proposed in this paper has a higher prediction accuracy.At the same time,this article also proposes a new solution idea and method for the forecast of financial time series.