Quantitative Stock Selection Model Based On Deep Learning

With the continuous improvement of big data,computing power and artificial intelligence technology,how to use the information in machine learning data has become the focus of people's attention.Deep learning is also applied to various fields under such circumstances,and the field of quantitative investment is also concerned as one of the key areas.Quantitative investment refers to the modeling and analysis of financial-defined numerical data to predict the investment income of financial products.Researching the application of deep learning in quantitative investment on the one hand is to verify the possibility of applying deep learning to quantitative investment.On the other hand,it is to help investors make more effective financial investments and obtain higher returns with full use of past information.At present,there are problems in the current application area.Firstly,selected features are not enough,and the prediction is prone to overfitting.Secondly,the model design is not reasonable,and the effect of the deep learning model is not fully exerted.Thirdly,different deep learning models have not been used in the quantitative investment analysis.This article mainly studies the application of deep learning models in quantitative investment analysis,using convolutional neural networks and long-term and short-term memory networks in deep learning to predict stock data.For convolutional neural network models,the data is processed from a large number of features to a small number of characteristic indicators,and construct a single sample according to the time dimension to analyze stock fluctuations.For the long-short term memory network model,the data is processed into a single sample,training stock to classify and forecast their ups and downs.Then a CNN + LSTM model will be constructed based on the combination of the two models,and the sample will be reprocessed to predict the rise and fall.Finally,the two basic models and the CNN+ LSTM model and the XGBoost timing + LSTM stock selection model are applied to stock backtesting to verify its application in the actual market and confirm its effectiveness.A comparative analysis of the effects of these four methods is given.The comparative analysis shows that the application of deep learning in the stock market is possible and that it has good performance.The innovations of this paper are: 1.Summarize the effectiveness of deep learning in quantified investment,and use the advantages of the model to build a CNN + LSTM model and obtain the best prediction effect;2.Select a large number of feature factors for training to extract as much hidden information as possible from the stock market;3.During the backtesting phase,combine timing and stock selection to build the XGBoost timing+ LSTM stock selection model,in order to get a better investment portfolio and obtain better income.