The Research Of Trading Strategies On SSE50 Derivatives Based On Investor Sentiment And Machine Learning

Among various fields of research and application about behavioral finance,those focus on investor sentiment have taken a considerable role in terms of amount of efforts poured into and impacts brought out,trading strategies based on investor sentiment have drawn extensive attention.Given the fact that China stock market is dominated by retail investors,we believe that major stock indices like SSE50 and their derivatives are more sensitive to the fluctuation of investor sentiment.This thesis is aimed to design trading strategies on SSE50 derivatives based on investor sentiment and improve it further using machine learning as an innovative tool.This thesis applies the method of Principal Component Analysis to extract common information from 8 investor sentiment measures and construct a comprehensive investor sentiment index,which has been proved to have ideal effectiveness.Trading strategies of SSE50 ETF and Index futures have been designed based on this comprehensive investor sentiment index using two different models – the traditional Vector Auto-regression model and the Nonlinear Auto-regressive Model with Exogenous Input(NARX)neural network model.The back-testing results and robustness analysis have revealed that the trading strategies based on investor sentiment significantly outperform the market when the market is irrational and highly volatile.While on the same time,machine learning method has been proved to be more effective when modeling and forecasting market price movement than the traditional econometric method.