A novel statistical and machine learning hybrid approach to predicting S&P500 using sentiment analysis

Tools

Murtagh, Fionn; Olaniyan, Rapheal and Stamate, Daniel. 2015. 'A novel statistical and machine learning hybrid approach to predicting S&P500 using sentiment analysis'. In: 8th International Conference of the ERCIM Working Group on Computational and Methodological Statistics. Senate House, University of London, United Kingdom. [Conference or Workshop Item]

Preview

Text
cms_2015.pdf
Download (32kB) | Preview

Official URL: http://www.cmstatistics.org/CMStatistics2015/

Abstract or Description

The frequent ups and downs are characteristic to the stock market. The conventional standard
models that assume that investors act rationally have not been able to capture the irregularities in
the stock market patterns for years. As a result, behavioural finance is embraced to attempt to
correct these model shortcomings by adding some factors to capture sentimental contagion which
may be at play in determining the stock market. The authors address the predictive influence of
online expressed sentiment on the stock market returns and volatility by using a non-parametric
nonlinear approach that corrects specific limitations encountered in previous approaches. A novel
approach to developing sentiment analysis and stock market predictive models based on GARCH,
EGARCH and recurrent neural network frameworks is presented, and is compared to previous
statistical and/or machine learning approaches addressing this problem, proving its advantages and
superiority over the latter. The sentiment information extracted via text mining from online blogs
includes variants of indexes expressing relevant sentiment, in particular anxiety, whose predictive
value on the dynamic of S&P 500 is rigorously analysed using linear and nonlinear Granger causality
and Monte Carlo simulations. Future extensions envisage incorporating the necessary apparatus and
efficient mechanism to handle also stream data.