Stock Market Prediction Based On Deep Learning And Technical Analysis

The analysis and prediction of stock market has always been the concern of researchers.Analysis methods usually include fundamental analysis and technical analysis.At the same time,with the rapid development of deep learning,the model of "technical analysis + deep learning" is widely used in academic research and practice.Therefore,this paper takes the technical indicators as the data characteristics,and carries out prediction and result evaluation through convolutional neural network.In this paper,taking the market conditions of constituent stocks of CSI 300 index,S & P500 index,FTSE 100 index,Nikkei 225 index,S&P/TSX index,CAC40 index,DAX30 index and ftmib index from 2011 to 2020 as the data set,the extracted data features include opening price 5 fundamental analysis indexes including closing price and 31 technical analysis indexes including Ma,SMA and RSI.The 36 data features of 36 consecutive days are formed into a two-dimensional gray image,and then marked with the rise and fall of the last day of 36 days.Take the images and labels from 2011 to 2015 as the training set and the images and labels from 2016 as the test set for training.In this way,push the training set and test set back one year at a time for training and testing.Then,the convolution neural network is used to identify the two-dimensional image,and then the bullish and bearish situation from 2016 to 2020 is obtained.Next,according to the predicted rise and fall labels,a simulated trading strategy is constructed,and the trading strategy obtained by CNN model is evaluated and compared with bah strategy,RSI strategy and SMA strategy.The results show that the accuracy of CNN model with technical indicators as data characteristics is about 70% in the prediction of index components in eight countries,including 72.56% in the prediction of Shanghai and Shenzhen 300 index components,which can better predict the rise and fall of stocks.In the simulation transaction evaluation stage,whether considering the transaction cost or not,CNN strategy performs best in terms of yield,risk and sharp index compared with the other three strategies.At the same time,it can be seen that due to different degrees of market effectiveness in 8 countries,China’s market performs well in terms of yield and maximum pullback.The results show that the accuracy of the CNN model in predicting the constituent stocks of the eight countries’ index is about 70%,which can better predict the rise and fall of stocks,and at the same time,it can test the differences between the markets of different countries.In the simulated transaction evaluation stage,regardless of whether transaction costs are considered,the CNN strategy performs best in terms of return,risk,and Sharpe index compared to the other three strategies.At the same time,it can be seen that the stock investment portfolio in a financial market with a certain scale has a higher rate of return.