Classification Based Stock Price Prediction Using Deep Residual Network

In the modern society's economic system and social organization,the financial market occupies a key position.Analyzing and predicting the behavior of financial market can provide a reference for investors to make investment plans and decisions,which not only improve the investment rationality and obtain higher profits,but also maintain the healthy development of financial market.Stock price forecasting is an important means to improve returns and reduce risks in stock investment,and it is also an important direction in financial time series research.In the past,much of this research has focused on linear time series models such as ARMA and GARCH.In recent years,with the development of machine learning,various types of neural networks are increasingly used in stock price prediction,such as SVM,DNN,and CNN.Most of these algorithms use numerical features such as original price series and technical indicators as model inputs.Considering that people often use stock price graphs as the basis for decision making,such as the "Head and Shoulders Bottom","cross stars" and so on,the graphical features may contain more abundant information that is difficult to express in numerical characteristics.Therefore,this paper proposes a new deep learning prediction model,using the stock price graph as input for the deep residual network(ResNet).This model makes full use of the feature extraction ability of the deep network in the image and improves the prediction accuracy.In addition,we implemented a stock T+0 trading strategy based on the forecast model and conduct simulation transactions on 10 China Market index stock data sets and 10 Baima stock data sets.This strategy makes full use of the rising and falling two-way forcasting signals.The results show that,"T+0" strategy gets more profits than BH strategy(Buy and Hold).On the same data sets,the profitability of the ResNet model based on graph features exceeds that of the SVM,DNN,and CNN models,which proves its efficiency in the field of stock price forecasting.