Forecast Of CSI 300 Index Based On Convolutional Neural Network

Since the beginning of the stock market,countless researchers and investors at home and abroad have been involved in the prediction of stock prices.Especially after the rise of artificial intelligence technology,the application of artificial intelligence technology to stock prediction has quickly become a research hotspot.This thesis makes a useful exploration on the combination of stock forecasting and artificial intelligence in actual projects.In the thesis,a convolutional neural network is set up through the neural network library Keras,and its supervised learning performance is applied to the prediction of ups and downs of the CSI 300 Index.And the forecasting result is good at last,which will be helpful for the subsequent investment strategy formulation.In the research,a variety of feature engineering methods were used to study the data,and the neural network model was adjusted from different perspectives to make the model better adapt to the financial data,which played a greater learning ability.This thesis firstly introduce the basic concepts of the stock market and the research status at home and abroad,and the convolution neural network method according to the actual project requirements.Secondly the thesis focuses on the financial data studied(CSI300 Index),and uses a variety of feature engineering methods to do research,among which there are some innovations in feature construction.Then in the thesis,the conception of convolution neural network model construction and tuning process is elaborated.The tuning process is an important part of the convolution neural network model.Through tuning the model from different perspectives,a good prediction accuracy is finally obtained.Finally,we compare the research results of this thesis with other machine learning methods,and summarize the thesis and looks forward to the prospects of the future research.