Research On Stock Forecasting Based On The Multi-view Feature Data

The research of stock forecasting is an applied research direction of financial big data.With the development of information technology,the stock forecasting research is not only rigidly adhered to the basic analysis methods,but also uses the technical analysis methods mostly,such as machine learning methods,and has achieved certain significant research results.Among them,the neural network method provides a new modeling method for stock forecasting research.Based on the related background,methods or models and related theoretical knowledge of the stock forecasting problem,this thesis introduces the concept of perspective into the research of the stock forecasting problem and studies the stock forecasting from the perspective.The main research works of this thesis has the following aspects:1.Propose a BPNN model based on single-view feature data,and conduct the stock forecasting research.From the perspective,using the BPNN model based on single-view feature data for stock forecasting research.Based on the stock historical data of the Shanghai Stock Exchange Shanghai Pudong Development Bank(Stock Code: SH600000)of the A-shares of China's stock market,after processed by the financial big data's basis data processing platform,the feature data from the perspective of the corresponding technical indicators are obtained,and conduct the empirical analysis of this model based on the single technical indicators view feature data,including optimal network structure and stock forecasting results.2.Propose a BPNN model based on information fusion of multi-view feature data,and conduct the stock forecasting research.To solve the incompleteness of information of single-view feature data,from the perspective of multi-view feature data,feature data of multiple views are fused at feature level through information fusion and the stock forecasting research is studied by using the BPNN model based on information fusion of multi-view feature data.Based on the stock historical data of the Shanghai Stock Exchange Shanghai Pudong Development Bank(Stock Code: SH600000)of the A-shares of China's stock market,after processed by the financial big data's basis data processing platform,the information fusion of multiple technical indicators view feature data are obtained,and conduct the empirical analysis of this model,including optimal network structure and stock forecastingresults.The experimental results show that the prediction results based on multi-view information fusion feature data are better than the prediction results based on single-view feature data,and are also better than the results of BPNN model based on stock historical data.3.Propose a BPNN model based on integrated learning of multi-view feature data,and conduct the stock forecasting research.At the same time,it also proposes the corresponding integrated learning algorithm,namely BagMVFD-BPNN algorithm.To solve the incompleteness of information of single-view feature data,from the perspective of multi-view feature data,by the way of integrated learning,using the BagMVFD-BPNN algorithm,the BPNN model based on integrated learning of multi-view feature data is used for stock forecasting research.Based on the stock historical data of the Shanghai Stock Exchange Shanghai Pudong Development Bank(Stock Code: SH600000)of the A-shares of China's stock market,after processed by the financial big data's basis data processing platform,the multiple technical indicators view feature data are obtained,and conduct the empirical analysis of this model,including optimal network structure and stock forecasting results.The experimental results show that the prediction results based on multi-view feature data integrated learning are better than the prediction results based on single-view feature data,and are superior to the results of the prediction results based on multi-view information fusion feature data,and are also better than the results of BPNN model based on stock historical data.At the same time,it is also known that the degree of the importance of each technical index perspective and the major technical indexes affecting the stock forecasting results,that is,the multi-view feature selection of multi-view feature data.Thus can provide investors with a reference in the perspective of major technical indexes and reference opinions on the prediction results for the stock investment of shareholders,and guide the stock investment of shareholders.