Research On Stock Market Prediction Via Multi-task And Feature Interaction

The investment on the stock market are prone to be affected by the Internet.With the advent of the Internet ero,a number of studies show that the financial news and posts in the social network have influences on the decisions of investors.Therefore,in addition to traditional quantitative data,public opinions in the the Internet are also required to considered to predict stock market fluctuation.For the purpose of improving the prediction accuracy,we propose a multi-task stock prediction model that not only consideres the stock correlations but also supports multi-source data fusion.Our proposed model first utilizes tensor to integrate the multi-source data,including financial Web news,investors' sentiments extracted from the social network and some quantitative data on stocks.In this way,the intrinsic relationships among different information sources can be captured,and meanwhile multi-sourced information can be complemented to solve the data sparsity problem.Secondly,a new metric to define the stock correlation based on the coupling effects among the features is proposed.The stock similarity measure considers both the similarity between different days for a specific stock and the similarity between different stocks for a day.This metric performs well on handling nominal attributes,and it also capture real-world relationships.Thirdly,we propose an improved sub-mode coordinate algorithm(SMC).SMC is based on the stock similarity,aiming to reduce the variance of their subspace in each dimension produced by the tensor decomposition.The algorithm is able to improve the quality of the input features,and thus improves the prediction accuracy.Fourthly,the paper utilizes the Long Short-Term Memory(LSTM)neural network model to predict the stock fluctuation trends.Finally,the experiments on seventy-eight A-share stocks in CSI100 and thirteen popular HK stocks in the year 2015 and 2016 are conducted.The results demonstrate the improvement on the prediction accuracy and the effectiveness of the proposed model.