Design And Implementation Of Real-time Stock Prediction System Based On Social Network Sentiments Analysis

The variation of stock market,is one of the most important parts of capital market fluctuations,is always expected to be predicted by investors.The prediction of its variation provides not only useful advice for investors,but also signals for economic development,which is beneficial to the stability and prosperity of the capital market.However,since the complete market information is always exposed to the public and the trade is free,stock prices often show their randomness,and predictions are always challenging.Nevertheless,not all the investors are rational enough,which makes that emotional tendency towards a specific company will definitely affect its market performances and prices in stock prices.Therefore,to some degree,stock price prediction in a short period is not impossible.To address this prediction problem,this thesis designs and implements a real-time stock prediction system based on online social network sentiments analysis algorithm.This system is divided into several modules:data access module,data transport module,data pre-process module,data calculation module,data storage module and data visualization module.By acquiring social network sentiment data towards specific listed companies,and transporting them via high availability network channel,the system can pre-process these data,and calculate them.This thesis proposes an improvement for the elimination strategy of an Online SVM algorithm,and proposes an Online Passive Aggressive SVM algorithm based on minimize single update step length.This thesis also implements stock price prediction based on time series algorithm which is applied in data calculation module.Finally,the system outputs and shows prediction results by accumulating sentiment data and history stock prices.This thesis first validates the availability of infrastructures of the stream system,then improves the accuracy of designed algorithm,finally validates the complete project in functional and non-functional test.