Research On Stock Linkage Effect Based On Heterogeneous Information Processing

In recent years,the emergence of computer finance has evaded the subjectivity and empiricism of financial analysts in controlling financial markets.Therefore,many Internet platforms have started to provide stock diagnosis,stock recommendation,quantitative investment and other services.China's major banks and some Internet companies have also entered the Fintech era of integration of finance and computer technology machine learning.The various wealth management solutions launched under the background of computer finance will become one of the major ways for people to invest in financial management in the future economic development.Aiming at the problem of stock linkage effect,based on the data set of financial heterogeneity,this paper generates the candidate series of linked stocks by calculating the similarity of the stock price fluctuation timing curves,and uses the machine learning algorithm to judge the joint stock candidate sets.The linkage result set generated by the linkage decision method aims to provide investors with auxiliary decision suggestions.This topic mainly studies from the following aspects:Financial heterogeneous data set construction.There are many influencing factors of stock linkage effects,which are mainly divided into unstructured financial text information and structured financial data information.Crawling and cleaning financial text information from financial websites(Sina Finance,Orient Fortune,Sohu Securities,etc.)through crawler technology to process and process structured text messages.In order to ensure the accuracy of the financial data set,cross-validated the Sina stock timing API,Yahoo Finance stock timing API and open source Tushare's stock timing API to get accurate financial data.Combining the processed textual information and data information to construct a heterogeneous dataset that supports the study of stock linkage effects.Generation of Linked Stock Candidate Sets.The stock linkage effect reflects the similarities and differences between the historical trading data on the characteristics of the stock data.The strength of the stock linkage effect is positively correlated with the size of the similar value.In this paper,we will compare the similarity calculation methods based on the same direction ratio,the absolute distance value and the Pearson correlation coefficient to generate the joint stock candidate set,which will be used as the input sample set for the linkage judgment method.Research on the Method of Stock linkage Judgment.By comparing SVM,SVR,and stock prediction performance under the input of multi-characteristic time series samples,the prediction of the method with better performance is used as the method of linkage judgment One of the sample eigenvalues.Combined with the constructedfinancial heterogeneous dataset,the stock linkage discriminant method based on machine learning algorithm is proposed.By using this method,the data regularity of the linked stock candidate set is analyzed and the stock linkage effect under different time differences is studied.The subject of this research is Shenzhen stock.By comparing the predictive performance of SVM,SVR and LR to the stock timing,we choose the predictive value output from the optimal forecasting algorithm and construct the eigenvalue of the sample of the linkage discriminant method with the stock heterogeneous eigenvalue.Based on the similarity of stock price,a series of candidate stock sets are generated,and compared with the stock-based stock selection method based on Xgboost,SVM and SVR to make linkage judgment.The linkage result set generated by the linkage decision method aims to provide investors with decision support advice.